Smitch The Unshakable Resolve of "this guy are sick" Member Apr 21, 2022 4,758 Mat Piscatella (@matpiscatella.bsky.social) April 2025 U.S. Video Game Market Highlights from Circana - April 2025 total market projected U.S. consumer spending across video game hardware, content and accessories fell 3% when compared to a year ago, to $4.1B. 2025 year-to-date spending was 8% behind 2024’s pace, at $17.8B. bsky.app Mat Piscatella (@matpiscatella.bsky.social) U.S. Video Game Market April Spending Trend - April video game content spending declined by 2% compared to a year ago, to $3.7B. 18% YoY growth in non-mobile subscription spend and a 16% gain in Console digital premium downloads were offset by declines in other content segments. bsky.app Mat Piscatella (@matpiscatella.bsky.social) U.S. Video Game Market 13-Month Spending Trend bsky.app View: https://bsky.app/profile/matpiscatella.bsky.social/post/3lpoo44vmwk2e View: https://bsky.app/profile/matpiscatella.bsky.social/post/3lpoo47q3hk2e Mat Piscatella (@matpiscatella.bsky.social) The Elder Scrolls IV: Oblivion: Remastered was the best-selling game of the month, instantly becoming the 3rd best-selling game of 2025 year-to-date. It was the best-selling title across each of the PlayStation, Xbox and PC platforms. bsky.app Mat Piscatella (@matpiscatella.bsky.social) The Elder Scrolls IV: Oblivion: Remastered sold more units in Apr 2025 than The Elder Scrolls IV: Oblivion achieved across its first 15 months in market combined following its March 2006 debut and generated more full game dollar sales than the original’s first 14 months combined. bsky.app Mat Piscatella (@matpiscatella.bsky.social) Forza Horizon 5 was the 2nd best-selling game of April 2025 following its debut on PlayStation 5. It ranked as the 42nd best-selling title of March. bsky.app Mat Piscatella (@matpiscatella.bsky.social) Indiana Jones and the Great Circle also launched on PlayStation 5 during April 2025. The game jumped from 118th on the monthly best-selling titles chart a month ago to 6th. bsky.app Mat Piscatella (@matpiscatella.bsky.social) Four of the top five - and five of the top seven - best-selling games of the month on PlayStation platforms were published by Microsoft. bsky.app Mat Piscatella (@matpiscatella.bsky.social) Call of Duty HQ led all titles in total US monthly active users across both PlayStation and Xbox platforms during April, according to Circana’s Player Engagement Tracker. Schedule 1 ranked 1st in US total month active users on Steam. bsky.app Mat Piscatella (@matpiscatella.bsky.social) Two titles saw strong engagement gains on PlayStation following their addition to PS+. RoboCop: Rogue City jumped from 111th in PlayStation total monthly active users in March to 8th, while The Texas Chain Saw Massacre jumped to #10 in April from 375th a month ago. bsky.app Mat Piscatella (@matpiscatella.bsky.social) April hardware spending fell by 8% when compared to a year ago, to $186M. It was the lowest monthly hardware dollar sales total reached in the US market since July 2020 ($163M). bsky.app Mat Piscatella (@matpiscatella.bsky.social) PlayStation 5 dollar sales declined 5% in April compared to a year ago, but the platform once again led the hardware market in both dollar and unit sales. Switch hardware spending fell 37% in April vs YA, while Xbox Series grew 8%. bsky.app Mat Piscatella (@matpiscatella.bsky.social) April spending on accessories fell by 2% when compared to a year ago, to $166 million. The PS5 Dual Sense Wireless Controller Midnight Black was April’s best-selling accessory in both unit and dollar sales. bsky.app   OP OP Smitch The Unshakable Resolve of "this guy are sick" Member Apr 21, 2022 4,758 Maelstrom Member Apr 22, 2025 151 Okay, seems like the same thing as last week. And the week before. And the week before..   Bigkrev Member Oct 25, 2017 13,000 Does Kepler not report information? This should have captured the first ~10 days of Clair Obscur Expedition 33, but it doesn't show up anywhere on these lists, including engagement, which theoretically would have covered if people just gamepass'd the game...   Derbel McDillet ▲ Legend ▲ Member Nov 23, 2022 24,351 Good for Hello Kitty Island Adventure.   jroc74 Member Oct 27, 2017 34,053 Hell of a jump for both Forza and Indians Jones. MLB The Show doing MLB The Show things.  Granjinhaa Member Dec 28, 2023 9,627 Bigkrev said: Does Kepler not report information? This should have captured the first ~10 days of Clair Obscur Expedition 33, but it doesn't show up anywhere on these lists, including engagement, which theoretically would have covered if people just gamepass'd the game... Click to expand... Click to shrink... don't think they do   cw_sasuke Member Oct 27, 2017 29,988 Glad to see The Hundred Lines in the Switch charts at least. Switch outsold Xbox ? Seems kinda wild with Switch 2 releasing soon and a kinda stacked release schedule for Xbox and GP. Edit. Clarified a couple posts below. Xbox is 2nd and Switch 3rd for this month.  Last edited: Yesterday at 9:31 AM Gavalanche Prophet of Regret Member Oct 21, 2021 25,917 Bigkrev said: Does Kepler not report information? This should have captured the first ~10 days of Clair Obscur Expedition 33, but it doesn't show up anywhere on these lists, including engagement, which theoretically would have covered if people just gamepass'd the game... Click to expand... Click to shrink... No they do not. They are an indie publisher, most of them don't.  turkoman_ Member Apr 29, 2023 520 I can understand Call of Duty, GTA or Minecraft but Spiderman always there is pretty impressive imo.   Splinky Member Jul 12, 2023 146 Bigkrev said: Does Kepler not report information? This should have captured the first ~10 days of Clair Obscur Expedition 33, but it doesn't show up anywhere on these lists, including engagement, which theoretically would have covered if people just gamepass'd the game... Click to expand... Click to shrink... Pretty sure they don't. Kind of a BG3 situation where an obvious major player is completely missing from Circana's data   Mr Swine The Fallen Oct 26, 2017 6,931 Sweden cw_sasuke said: Glad to see The Hundred Lines in the Switch charts at least. Switch outsold Xbox ? Seems kinda wild with Switch 2 releasing soon and a kinda stacked release schedule for Xbox and GP. Click to expand... Click to shrink... One would think that Xbox would start outselling Switch before Switch 2 hits the market  jroc74 Member Oct 27, 2017 34,053 cw_sasuke said: Glad to see The Hundred Lines in the Switch charts at least. Switch outsold Xbox ? Seems kinda wild with Switch 2 releasing soon and a kinda stacked release schedule for Xbox and GP. Click to expand... Click to shrink... What remains shocking to me is Xbox can sell better, and still come in 3rd.   Granjinhaa Member Dec 28, 2023 9,627 cw_sasuke said: Switch outsold Xbox ? Seems kinda wild with Switch 2 releasing soon and a kinda stacked release schedule for Xbox and GP. Click to expand... Click to shrink... it did? i don't see in the bluesky thread   Threadmarks Xbox #2, Switch#3 New Index Tsunami561 Member Mar 7, 2023 5,359 cw_sasuke said: Switch outsold Xbox ? Seems kinda wild with Switch 2 releasing soon and a kinda stacked release schedule for Xbox and GP. Click to expand... Click to shrink... No View: https://bsky.app/profile/matpiscatella.bsky.social/post/3lpootbqmxh2e  New Index MarcosBrXD Member Aug 28, 2024 1,703 [CITAÇÃO="Sr. Swine, postagem: 140267553, membro: 8340"] Alguém poderia pensar que o Xbox começaria a vender mais que o Switch antes do Switch 2 chegar ao mercado [/CITAR] Principalmente nos EUA  Lant_War Classic Anus Game The Fallen Jul 14, 2018 25,286 Half the top 10 of PS is Microsoft Corp. Crazy   YozoraXV Member Oct 30, 2017 3,799 Real MS domination, interesting how ESO also got a boost from Oblivion.   Fabs Member Aug 22, 2019 2,753 Project Latitude in full effect. Xbox killing it right now. Bad Hardware year for everyone continues. Switch 2 can't come soon enough.  Mr Swine The Fallen Oct 26, 2017 6,931 Sweden Ah I see that it did outsell Switch, well good for Microsoft   FarSight XR-20 Member Jan 4, 2018 9,493 Oblivion Remastered sold more in 8 days than original game in first 15 months. The Elder Scrolls IV: Oblivion: Remastered sold more units in Apr 2025 than The Elder Scrolls IV: Oblivion achieved across its first 15 months in market combined following its March 2006 debut and generated more full game dollar sales than the original's first 14 months combined. Click to expand... Click to shrink...   mrmickfran The Fallen Oct 27, 2017 33,202 Gongaga Holy fuck @ Forza Hundred Line Defense in top 5 Switch titles is something? I guess  Terbinator Member Oct 29, 2017 13,373 XBS(+8%) Click to expand... Click to shrink... All according to keikaku?   UraMallas Member Nov 1, 2017 24,464 United States Xbox HARDWARE up? Well, you love to see that.  Maelstrom Member Apr 22, 2025 151 Lant_War said: Half the top 10 of PS is Microsoft Corp. Crazy Click to expand... Click to shrink... It won't be soon.   Tasman1991 Member Oct 22, 2024 446 Fabs said: Project Latitude in full effect. Xbox killing it right now. Bad Hardware year for everyone continues. Switch 2 can't come soon enough. Click to expand... Click to shrink... Bro PlayStation is down 5% and Microsoft's up 8% how is this bad   Mr Evil 37 Member Mar 7, 2022 27,772 YozoraXV said: interesting how ESO also got a boost from Oblivion. Click to expand... Click to shrink... It also had a showcase and updates and stuff. Don't think it was just Oblivion.   JAY_HORROR Member Sep 27, 2021 15,687 jroc74 Member Oct 27, 2017 34,053 Ok, this is better. Yeah....that price increase is gonna be rough.  Fabs Member Aug 22, 2019 2,753 Tasman1991 said: Bro PlayStation is down 5% and Microsoft's up 8% how is this bad Click to expand... Click to shrink... YOY and last April was bad. Lowest hardware spending since the covid shortages when everything is in stock and expensive.  GalvoAg Member Oct 30, 2017 2,608 Dallas Tasman1991 said: Bro PlayStation is down 5% and Microsoft's up 8% how is this bad Click to expand... Click to shrink... For real seems pretty healthy if anything considering Switch 2 around the corner.   Tasman1991 Member Oct 22, 2024 446 jroc74 said: Ok, this is better. Yeah....that price increase is gonna be rough. Click to expand... Click to shrink... So you're thinking everyone's getting their Xbox before are the price increase hits honestly I'm shocked that PlayStation 5 it's only - 5% down from last year that is insane   Gavalanche Prophet of Regret Member Oct 21, 2021 25,917 It makes sense there would be an increase on xbox consoles, everyone getting in before they increase the price. Maybe it explains why ps5 only dropped 5% as well, as the expectation is that they will also have a price increase.   Killer Member Oct 27, 2017 2,976 Microsoft PlayStation   g-m1n1 Member Oct 27, 2017 2,870 Luxembourg PS5 is Xbox Game Studios main platform. What a twist!  Gavalanche Prophet of Regret Member Oct 21, 2021 25,917 g-m1n1 said: PS5 is Xbox Game Studios main platform. What a twist! Click to expand... Click to shrink... They did release like six games this April, it is pretty expected :P  Tasman1991 Member Oct 22, 2024 446 Fabs said: YOY and last April was bad. Lowest hardware spending since the covid shortages when everything is in stock and expensive. Click to expand... Click to shrink... You do realise April's a quiet month right like you're not going to have constant high sales after it's peak right like you do realize this it will come down I don't want to be mean but it seems as you just go into these form posts and go this is bad and then just leave yet very 5% down from last year is actually very good for PlayStation considering they'd at at 2023 also at least Xbox is up in the Home Country through that could also be because people are getting Xbox is before the price increase   YupiScroopy Member Jun 28, 2023 149 LETS GO HUNDRED LINE!!  Mr Evil 37 Member Mar 7, 2022 27,772 g-m1n1 said: PS5 is Xbox Game Studios main platform. What a twist! Click to expand... Click to shrink... Not really if you look at the Doom numbers lol (or the Forza numbers). Sales might be high on PS but overall playerbase is still higher on Xbox.   mietek Member May 9, 2025 37 Gavalanche said: It makes sense there would be an increase on xbox consoles, everyone getting in before they increase the price. Click to expand... Click to shrink... The price hike was announced effective immediately, let's not revise history.   DamageEX2 Member May 20, 2024 1,247 No Expedition 33?   mietek Member May 9, 2025 37 Killer said: Microsoft PlayStation Click to expand... Click to shrink... Xbox Series P   cw_sasuke Member Oct 27, 2017 29,988 Thanks, this makes much more sense. The title made it seems like Switch was #2.  Miyoshi Member May 4, 2025 32 Lant_War said: Half the top 10 of PS is Microsoft Corp. Crazy Click to expand... Click to shrink... Yes, that's precisely the result of acquiring two major publishers and branding them under your name.   Mr Evil 37 Member Mar 7, 2022 27,772 RPGamer92 Member Oct 25, 2017 5,147 That's very good for FH5 especially as it came out at the end of the month.   g-m1n1 Member Oct 27, 2017 2,870 Luxembourg Mr Evil 37 said: Not really if you look at the Doom numbers lol (or the Forza numbers). Sales might be high on PS but overall playerbase is still higher on Xbox. Click to expand... Click to shrink... Of course talking about direct sales only (XGP don't show up on Circana numbers, except "player engagement"). If someone said this 3 years ago, he would be mocked by the whole forum.  Fabs Member Aug 22, 2019 2,753 Tasman1991 said: You do realise April's a quiet month right like you're not going to have constant high sales after it's peak right like you do realize this it will come down I don't want to be mean but it seems as you just go into these form posts and go this is bad and then just leave yet very 5% down from last year is actually very good for PlayStation considering they'd at at 2023 also at least Xbox is up in the Home Country through that could also be because people are getting Xbox is before the price increase Click to expand... Click to shrink... Both things can be true, and you would assume people would try to be purchasing hardware before Tariffs. April is historically bad, but dollars sales is alarming because a 700$ Pro exists. April 2023 was bad and this is worse. The whole year is down and likely to go down further on the big boy consoles with price increases. Switch 2 is needed just to stabilize the market and GTA being a year away is going to be a long wait.  Shopolic Avenger Oct 27, 2017 8,093 Forza Horizon 5! :O Imagine FH6 sales number when it'll be available day one on PS5!  ProdigyZA Member Jun 9, 2024 1,510 g-m1n1 said: PS5 is Xbox Game Studios main platform. What a twist! Click to expand... Click to shrink... This is not true.  

Bungie's Marathon reboot has been under fire for plagiarism of a British artist, and a new report has suggested Sony has set a sales target for the game that could prove challengingTech13:07, 21 May 2025Is Marathon in trouble?Destiny 2 developer Bungie isn't having the best of weeks, despite revealing a series of expansions for the game and preparing for the launch of its Marathon reboot in September.Still, the studio was forced to go back to the literal drawing board in recent days, with an artist claiming much of the game's design language was lifted from her portfolio, including some lifted wholesale.‌Now, Forbes' Paul Tassi, a regular reporter on Bungie, has discussed the high stakes facing the studio. Reports had previously suggested that Bungie is close to being absorbed into Sony and losing its autonomy if it underperforms, and that seems more likely than ever, given the publisher apparently wants the game to hit a high sales target.‌Posting the above video on YouTube where he explains the deep issues with Marathon on multiple levels, and how it's unlikely to be a critical hit, Tassi reveals he's heard Sony is expecting big things."I heard this needed to be in the top 5 of NPD sales for the year if it's going to be a success, or considered a success," he explained.Article continues below"Crazy," he adds, "there's no way".Last year, Sony did have a live service game in the top five NPD (now called Circana) games with Helldivers 2, but that felt like a cultural zeitgeist moment that didn't enter the arena with all of the baggage Marathon now has saddled on it.The gaming community praised Marathon's visuals, but now those are being questioned, too‌Considering 2025 has already seen critical hits like Monster Hunter Wilds and Split Fiction doing well sales-wise, and the Nintendo Switch 2 is launching soon, it seems like a tall mountain for Bungie to climb–especially since it's launching at a $40 price point when many alternatives in the genre are free-to-play.Still, it would be interesting for Sony to push a specific sales number, given that a live service title tends to live or die based on player retention.Over the weekend, Tassi reported that morale at the studio has plummeted.Article continues belowHe said: "Sony and Bungie legal are now sorting through this, and there is unlikely to be any new information as all of this continues to unfold. It is not clear how long an “audit” will take of the assets to remove or find any more plagiarism, as it’s an expansive enough process to have Bungie not even show any footage at all in its recent livestream.""Morale is in 'free-fall” across all departments, and “the vibes have never been worse.” Everyone has the same concerns about what happens to Bungie as a studio if Marathon bombs, which is something they absolutely cannot afford."For the latest breaking news and stories from across the globe from the Daily Star, sign up for our newsletters.‌‌‌

Got a project that’s too contemporary for your client? Submit your conceptual works, images and ideas for global recognition and print publication in the 2025 Vision Awards! The clock is ticking — submit your work ahead of the Main Entry deadline on June 6th. Spiral staircases are famously beautiful but notoriously tricky to design, particularly in section. (Ask any architect who’s ever tried!) Still, the payoff is worth it. Beyond their visual appeal, spiral stairs can save space, guide circulation, and even become striking centerpieces within a building. The projects below prove there’s more than one way to embrace the twist, using unexpected materials and creative details to shape spaces in offices, schools, restaurants and parks. Even when things get complicated, these nine stairs make a strong case for taking the spiral route. OFFICE @ 63 By Sanjay Puri Architects, Navi Mumbai, India Popular Choice Winner, 12th Annual A+Awards, Commercial Interiors (<25,000 sq ft.) A curved steel staircase rises through the center of this workplace, connecting three levels in one continuous movement. Perforated and matte-finished, it feels both solid and light. Its underside, lined with warm tones, softens the otherwise industrial palette. Set within a tall, multipurpose space used for exhibitions and gatherings, the stair becomes a quiet focal point. Surrounded by concrete, cork, glass and exposed metal, it stands out not just as a route between floors, but as an element that gives the entire space rhythm and focus. La Maison de Beauté Carita, L’Oréal-Luxe By Le studio REV, Paris, France At the center of Carita’s Paris beauty house, part of the L’Oréal portfolio, a curved staircase rises through what was once a closed courtyard. The space is now topped with a tall glass roof that brings in soft, natural light. Pale marble, pink onyx and brushed metal set a quiet, polished tone. The staircase is kept light with floating steps and slim supports. It connects the levels clearly and efficiently, without blocking views through arched openings. In a building designed around care and detail, the stair adds just the right touch of structure and calm movement. Sjustjärnan, New Nordic Headquarters for E.On By Kanozi Arkitekter, Malmö, Sweden A wide, open staircase connects the floors of E.ON’s Malmö headquarters, rising through a central glass atrium that joins the two curved volumes of the building. Designed to encourage movement and connection between departments, the stair is placed at the heart of shared circulation. Materials are simple and natural: light wood treads, glass balustrades and metal accents, chosen to support the building’s focus on well-being and daylight. Generous landings allow for pauses and casual interaction, reinforcing the workplace’s flexible layout. With no assigned desks and no divided entrances, the building encourages a sense of togetherness and the staircase plays a clear role in making that daily flow visible and accessible. The Learning Center at Quest By KSM Architecture, Chennai, India Jury Winner, 12th Annual A+Awards, Architecture +Learning A vivid red spiral staircase rises along the facade of this learning center in Chennai, connecting the top floor library to a rooftop cafeteria and terrace. Attached to the exterior, the stair adds a moment of surprise to the otherwise calm, ribbed concrete surface. Made from perforated metal, its bright finish contrasts with the muted palette of the building while echoing the colorful window frames scattered across the elevation. Inside, a separate metal stair winds through a central atrium, linking five stacked classrooms and open learning spaces. Together, the staircases reflect the school’s focus on movement, openness and curiosity — key principles in its interest-based approach to education. Varee Valley Restaurant : Jungle Junction By NPDA studio, Thailand Inside this forest-side restaurant in Northeast Thailand, a white spiral staircase rises slowly through the center of a light-filled dining area. Designed for pause and perspective, the stair leads visitors to a rooftop platform that opens up to the surrounding trees. Built with slender steel members and a delicate footprint, it reflects the lightness of the overall structure and mirrors the verticality of nearby tree trunks. The stair wraps around a central cylindrical column that also serves as a rainwater drain and light shaft, drawing daylight into the heart of the space. The Opera Park By Cobe, Copenhagen, Denmark Jury Winner, Public Parks & Green Spaces, 12th Annual A+Awards At the heart of this new harbor front park in Copenhagen, a spiral staircase descends through the center of a circular greenhouse, wrapping gently around a subtropical garden. The stair connects the café at park level with an underground parking facility, guiding visitors down through layers of greenery. Slim metal railings trace its curve, while the surrounding glass enclosure creates the feeling of walking through an open-air terrarium. The park itself spans six themed gardens across a former industrial island, offering paths, ponds and planting from around the world. In this setting, the staircase becomes part of a wider idea: architecture that supports slow movement, seasonal change and a close connection to nature. Balmy Palmy By CplusC Architects + Builders, Sydney, Australia In this compact holiday home raised above a rocky slope, a yellow spiral staircase links the natural ground to the open-air platform above. Built from powder-coated steel, it curves tightly between tree trunks and timber posts, bringing a sharp contrast to the greens and browns of the site. The stair leads into an outdoor area that connects the home’s simple rooms, with views of the treetops, the bay and the sky shifting between the leaves. Designed to limit excavation and preserve the forest setting, the staircase plays a clear role in the project’s minimal footprint — serving as a key access point and everyday reminder of the landscape it rises from. Dental Clinic By IFAgroup, Gdańsk, Poland Jury Winner, Hospitals and Healthcare Centers, 12th Annual A+Awards Jury Winner,  Architecture +Health, 12th Annual A+Awards Photos by Hanna Połczyńska This dental clinic, once a granary, is organized around a wide spiral staircase that rises through its three-story foyer. Clad in warm wood, the stair connects waiting areas, dental offices and training spaces while giving the interior a clear center. It’s surrounded by concrete, brick and steel surfaces that reflect the building’s industrial past. Live piano music and soft lighting help create a calm atmosphere, while sound-absorbing materials keep the space quiet and focused. Envisioned as more than a circulation route, the stair shapes how people move, wait and gather inside. Concept WRRF Yixing (Water Resource Recovery Facility) By SUP Atelier of THAD, China Inside this water treatment facility, a spiral staircase sits next to a café and community area. Built from metal with clean, simple details, it leads visitors through spaces designed to educate the public about recycling and sustainability. Large windows around the stair offer clear views of the nearby farmland and waterways, connecting visitors with the natural setting. As part of the facility’s open, welcoming design, the staircase makes it easy to see and understand how wastewater is turned into clean water and renewable energy. Got a project that’s too contemporary for your client? Submit your conceptual works, images and ideas for global recognition and print publication in the 2025 Vision Awards! The clock is ticking — submit your work ahead of the Main Entry deadline on June 6th. Top image: The Learning Center at Quest by KSM Architecture, Chennai, India The post 9 (Somewhat) Unconventional Takes on The Beloved Spiral Staircase appeared first on Journal.

Scientific publication T. M. Lange, M. Gültas, A. O. Schmitt & F. Heinrich (2025). optRF: Optimising random forest stability by determining the optimal number of trees. BMC bioinformatics, 26(1), 95.Follow this LINK to the original publication. Random Forest — A Powerful Tool for Anyone Working With Data What is Random Forest? Have you ever wished you could make better decisions using data — like predicting the risk of diseases, crop yields, or spotting patterns in customer behavior? That’s where machine learning comes in and one of the most accessible and powerful tools in this field is something called Random Forest. So why is random forest so popular? For one, it’s incredibly flexible. It works well with many types of data whether numbers, categories, or both. It’s also widely used in many fields — from predicting patient outcomes in healthcare to detecting fraud in finance, from improving shopping experiences online to optimising agricultural practices. Despite the name, random forest has nothing to do with trees in a forest — but it does use something called Decision Trees to make smart predictions. You can think of a decision tree as a flowchart that guides a series of yes/no questions based on the data you give it. A random forest creates a whole bunch of these trees (hence the “forest”), each slightly different, and then combines their results to make one final decision. It’s a bit like asking a group of experts for their opinion and then going with the majority vote. But until recently, one question was unanswered: How many decision trees do I actually need? If each decision tree can lead to different results, averaging many trees would lead to better and more reliable results. But how many are enough? Luckily, the optRF package answers this question! So let’s have a look at how to optimise Random Forest for predictions and variable selection! Making Predictions with Random Forests To optimise and to use random forest for making predictions, we can use the open-source statistics programme R. Once we open R, we have to install the two R packages “ranger” which allows to use random forests in R and “optRF” to optimise random forests. Both packages are open-source and available via the official R repository CRAN. In order to install and load these packages, the following lines of R code can be run: > install.packages(“ranger”) > install.packages(“optRF”) > library(ranger) > library(optRF) Now that the packages are installed and loaded into the library, we can use the functions that these packages contain. Furthermore, we can also use the data set included in the optRF package which is free to use under the GPL license (just as the optRF package itself). This data set called SNPdata contains in the first column the yield of 250 wheat plants as well as 5000 genomic markers (so called single nucleotide polymorphisms or SNPs) that can contain either the value 0 or 2. > SNPdata[1:5,1:5] Yield SNP_0001 SNP_0002 SNP_0003 SNP_0004 ID_001 670.7588 0 0 0 0 ID_002 542.5611 0 2 0 0 ID_003 591.6631 2 2 0 2 ID_004 476.3727 0 0 0 0 ID_005 635.9814 2 2 0 2 This data set is an example for genomic data and can be used for genomic prediction which is a very important tool for breeding high-yielding crops and, thus, to fight world hunger. The idea is to predict the yield of crops using genomic markers. And exactly for this purpose, random forest can be used! That means that a random forest model is used to describe the relationship between the yield and the genomic markers. Afterwards, we can predict the yield of wheat plants where we only have genomic markers. Therefore, let’s imagine that we have 200 wheat plants where we know the yield and the genomic markers. This is the so-called training data set. Let’s further assume that we have 50 wheat plants where we know the genomic markers but not their yield. This is the so-called test data set. Thus, we separate the data frame SNPdata so that the first 200 rows are saved as training and the last 50 rows without their yield are saved as test data: > Training = SNPdata[1:200,] > Test = SNPdata[201:250,-1] With these data sets, we can now have a look at how to make predictions using random forests! First, we got to calculate the optimal number of trees for random forest. Since we want to make predictions, we use the function opt_prediction from the optRF package. Into this function we have to insert the response from the training data set (in this case the yield), the predictors from the training data set (in this case the genomic markers), and the predictors from the test data set. Before we run this function, we can use the set.seed function to ensure reproducibility even though this is not necessary (we will see later why reproducibility is an issue here): > set.seed(123) > optRF_result = opt_prediction(y = Training[,1], + X = Training[,-1], + X_Test = Test) Recommended number of trees: 19000 All the results from the opt_prediction function are now saved in the object optRF_result, however, the most important information was already printed in the console: For this data set, we should use 19,000 trees. With this information, we can now use random forest to make predictions. Therefore, we use the ranger function to derive a random forest model that describes the relationship between the genomic markers and the yield in the training data set. Also here, we have to insert the response in the y argument and the predictors in the x argument. Furthermore, we can set the write.forest argument to be TRUE and we can insert the optimal number of trees in the num.trees argument: > RF_model = ranger(y = Training[,1], x = Training[,-1], + write.forest = TRUE, num.trees = 19000) And that’s it! The object RF_model contains the random forest model that describes the relationship between the genomic markers and the yield. With this model, we can now predict the yield for the 50 plants in the test data set where we have the genomic markers but we don’t know the yield: > predictions = predict(RF_model, data=Test)$predictions > predicted_Test = data.frame(ID = row.names(Test), predicted_yield = predictions) The data frame predicted_Test now contains the IDs of the wheat plants together with their predicted yield: > head(predicted_Test) ID predicted_yield ID_201 593.6063 ID_202 596.8615 ID_203 591.3695 ID_204 589.3909 ID_205 599.5155 ID_206 608.1031 Variable Selection with Random Forests A different approach to analysing such a data set would be to find out which variables are most important to predict the response. In this case, the question would be which genomic markers are most important to predict the yield. Also this can be done with random forests! If we tackle such a task, we don’t need a training and a test data set. We can simply use the entire data set SNPdata and see which of the variables are the most important ones. But before we do that, we should again determine the optimal number of trees using the optRF package. Since we are insterested in calculating the variable importance, we use the function opt_importance: > set.seed(123) > optRF_result = opt_importance(y=SNPdata[,1], + X=SNPdata[,-1]) Recommended number of trees: 40000 One can see that the optimal number of trees is now higher than it was for predictions. This is actually often the case. However, with this number of trees, we can now use the ranger function to calculate the importance of the variables. Therefore, we use the ranger function as before but we change the number of trees in the num.trees argument to 40,000 and we set the importance argument to “permutation” (other options are “impurity” and “impurity_corrected”).  > set.seed(123) > RF_model = ranger(y=SNPdata[,1], x=SNPdata[,-1], + write.forest = TRUE, num.trees = 40000, + importance="permutation") > D_VI = data.frame(variable = names(SNPdata)[-1], + importance = RF_model$variable.importance) > D_VI = D_VI[order(D_VI$importance, decreasing=TRUE),] The data frame D_VI now contains all the variables, thus, all the genomic markers, and next to it, their importance. Also, we have directly ordered this data frame so that the most important markers are on the top and the least important markers are at the bottom of this data frame. Which means that we can have a look at the most important variables using the head function: > head(D_VI) variable importance SNP_0020 45.75302 SNP_0004 38.65594 SNP_0019 36.81254 SNP_0050 34.56292 SNP_0033 30.47347 SNP_0043 28.54312 And that’s it! We have used random forest to make predictions and to estimate the most important variables in a data set. Furthermore, we have optimised random forest using the optRF package! Why Do We Need Optimisation? Now that we’ve seen how easy it is to use random forest and how quickly it can be optimised, it’s time to take a closer look at what’s happening behind the scenes. Specifically, we’ll explore how random forest works and why the results might change from one run to another. To do this, we’ll use random forest to calculate the importance of each genomic marker but instead of optimising the number of trees beforehand, we’ll stick with the default settings in the ranger function. By default, ranger uses 500 decision trees. Let’s try it out: > set.seed(123) > RF_model = ranger(y=SNPdata[,1], x=SNPdata[,-1], + write.forest = TRUE, importance="permutation") > D_VI = data.frame(variable = names(SNPdata)[-1], + importance = RF_model$variable.importance) > D_VI = D_VI[order(D_VI$importance, decreasing=TRUE),] > head(D_VI) variable importance SNP_0020 80.22909 SNP_0019 60.37387 SNP_0043 50.52367 SNP_0005 43.47999 SNP_0034 38.52494 SNP_0015 34.88654 As expected, everything runs smoothly — and quickly! In fact, this run was significantly faster than when we previously used 40,000 trees. But what happens if we run the exact same code again but this time with a different seed? > set.seed(321) > RF_model2 = ranger(y=SNPdata[,1], x=SNPdata[,-1], + write.forest = TRUE, importance="permutation") > D_VI2 = data.frame(variable = names(SNPdata)[-1], + importance = RF_model2$variable.importance) > D_VI2 = D_VI2[order(D_VI2$importance, decreasing=TRUE),] > head(D_VI2) variable importance SNP_0050 60.64051 SNP_0043 58.59175 SNP_0033 52.15701 SNP_0020 51.10561 SNP_0015 34.86162 SNP_0019 34.21317 Once again, everything appears to work fine but take a closer look at the results. In the first run, SNP_0020 had the highest importance score at 80.23, but in the second run, SNP_0050 takes the top spot and SNP_0020 drops to the fourth place with a much lower importance score of 51.11. That’s a significant shift! So what changed? The answer lies in something called non-determinism. Random forest, as the name suggests, involves a lot of randomness: it randomly selects data samples and subsets of variables at various points during training. This randomness helps prevent overfitting but it also means that results can vary slightly each time you run the algorithm — even with the exact same data set. That’s where the set.seed() function comes in. It acts like a bookmark in a shuffled deck of cards. By setting the same seed, you ensure that the random choices made by the algorithm follow the same sequence every time you run the code. But when you change the seed, you’re effectively changing the random path the algorithm follows. That’s why, in our example, the most important genomic markers came out differently in each run. This behavior — where the same process can yield different results due to internal randomness — is a classic example of non-determinism in machine learning. Taming the Randomness in Random Forests As we just saw, random forest models can produce slightly different results every time you run them even when using the same data due to the algorithm’s built-in randomness. So, how can we reduce this randomness and make our results more stable? One of the simplest and most effective ways is to increase the number of trees. Each tree in a random forest is trained on a random subset of the data and variables, so the more trees we add, the better the model can “average out” the noise caused by individual trees. Think of it like asking 10 people for their opinion versus asking 1,000 — you’re more likely to get a reliable answer from the larger group. With more trees, the model’s predictions and variable importance rankings tend to become more stable and reproducible even without setting a specific seed. In other words, adding more trees helps to tame the randomness. However, there’s a catch. More trees also mean more computation time. Training a random forest with 500 trees might take a few seconds but training one with 40,000 trees could take several minutes or more, depending on the size of your data set and your computer’s performance. However, the relationship between the stability and the computation time of random forest is non-linear. While going from 500 to 1,000 trees can significantly improve stability, going from 5,000 to 10,000 trees might only provide a tiny improvement in stability while doubling the computation time. At some point, you hit a plateau where adding more trees gives diminishing returns — you pay more in computation time but gain very little in stability. That’s why it’s essential to find the right balance: Enough trees to ensure stable results but not so many that your analysis becomes unnecessarily slow. And this is exactly what the optRF package does: it analyses the relationship between the stability and the number of trees in random forests and uses this relationship to determine the optimal number of trees that leads to stable results and beyond which adding more trees would unnecessarily increase the computation time. Above, we have already used the opt_importance function and saved the results as optRF_result. This object contains the information about the optimal number of trees but it also contains information about the relationship between the stability and the number of trees. Using the plot_stability function, we can visualise this relationship. Therefore, we have to insert the name of the optRF object, which measure we are interested in (here, we are interested in the “importance”), the interval we want to visualise on the X axis, and if the recommended number of trees should be added: > plot_stability(optRF_result, measure="importance", + from=0, to=50000, add_recommendation=FALSE) The output of the plot_stability function visualises the stability of random forest depending on the number of decision trees This plot clearly shows the non-linear relationship between stability and the number of trees. With 500 trees, random forest only leads to a stability of around 0.2 which explains why the results changed drastically when repeating random forest after setting a different seed. With the recommended 40,000 trees, however, the stability is near 1 (which indicates a perfect stability). Adding more than 40,000 trees would get the stability further to 1 but this increase would be only very small while the computation time would further increase. That is why 40,000 trees indicate the optimal number of trees for this data set. The Takeaway: Optimise Random Forest to Get the Most of It Random forest is a powerful ally for anyone working with data — whether you’re a researcher, analyst, student, or data scientist. It’s easy to use, remarkably flexible, and highly effective across a wide range of applications. But like any tool, using it well means understanding what’s happening under the hood. In this post, we’ve uncovered one of its hidden quirks: The randomness that makes it strong can also make it unstable if not carefully managed. Fortunately, with the optRF package, we can strike the perfect balance between stability and performance, ensuring we get reliable results without wasting computational resources. Whether you’re working in genomics, medicine, economics, agriculture, or any other data-rich field, mastering this balance will help you make smarter, more confident decisions based on your data. The post How to Set the Number of Trees in Random Forest appeared first on Towards Data Science.

To mark the 100th anniversary of McVitie's Chocolate Digestives, TBWA has launched a tongue-in-cheek campaign that positions the beloved snack alongside the wheel and the telephone in the canon of history's greatest human achievements. And frankly, they make a pretty compelling case. The campaign, which rolled out across out-of-home, digital, experiential and PR, takes a playful but purposeful approach to celebrating the "Choc Dig" – as the team affectionately call it – as a cultural icon. "The brief was to reassert the 'Choc Dig' as an irreplaceable British icon, using its 100th anniversary to elevate it from a mass brand to a timeless one," says Matt Tassell, Creative at TBWA. "We quickly realised that – despite its humble appearance and familiar reputation – the Choc Dig shares a number of traits with some of humanity's greatest inventions." It's a wry, knowing comparison, but one that's underpinned by genuine admiration for the biscuit's staying power and cultural significance. "The joke is we're not joking," Matt explains. "There's a fun tension at play in the campaign, and the lighthearted tone of voice grounds the reverential visuals." This tension plays out beautifully in the creative. Posters feature a giant Chocolate Digestive hovering above landscapes like some sort of snack-shaped Stonehenge, while a limited-run pop-up – The McVitie's Chocolate Digestives Experience – invited the public to indulge their inner biscuit nerd at London's Piccadilly Lights. "Our iconic poster campaign featured a Choc Dig looming large above all else, so it was only fitting that we ran it on Europe's largest single digital billboard," says Matt. "The fact that the Experience was hosted at the very base of the Lights, in the very epicentre of the nation's capital, tied everything together in celebratory fashion and reinforced the magnitude of the Chocolate Digestive's place in British culture." Of course, anniversaries like this are catnip for nostalgic Brits (myself included), but what elevates this campaign is the level of craft, humour and genuine storytelling across the full ecosystem. As Matt puts it, "We had a genuinely legendary product to play with and a simple, strong creative idea to anchor everything back to." That clarity of vision allowed TBWA to deliver a campaign that feels cohesive across channels while also giving collaborators and media partners a clear framework with which to align. At the heart of that framework is a distinct tone of voice that's affectionate, cheeky and quietly profound. It reminds people why they've loved Chocolate Digestives for so long but also hints at why they might continue to do so for the next hundred years. "We wanted to awaken some of that nostalgia and comfort," says Matt. "And playfully articulate its place in the world we all live in." That articulation went far beyond traditional media into a series of large-scale activations, including DOOH takeovers and landmark projections, which brought the campaign to life in ways that felt both irreverent and ambitious. "Bringing the DOOH and landmark projections to life meant juggling our big idea with even bigger tech," Matt admits. "There was plenty of discussion around timing, permissions, pixel capabilities, and, of course, the weather (how very British). But the payoff was more than worth it." The use of cutting-edge digital media also speaks to the other narrative strand running through the campaign – one of innovation and reinvention. "The great thing about the Chocolate Digestive is that its longevity doesn't define it exclusively," says Matt. "It's always on the front foot, pushing out new NPDs and fresh twists on its iconic blueprint." In other words, it's not just a relic of biscuit history – it's still a living, evolving part of British food culture. That's where this campaign really shines: it doesn't just mark 100 years of something old; it positions McVitie's as a brand that understands how to stay relevant. "Like any good innovator, McVitie's is ever-curious," Matt says. "Constantly developing new products, flavours, and collaborations to win over the next generation of biscuit munchers. It's all about staying as relevant and pivotal to British culture today as it has been for the past century." There's a lot to be said for a campaign that can simultaneously land a punchline, pull off a Piccadilly takeover, and plant the seed for future brand evolution. By treating the Chocolate Digestive with the same reverence (and irreverence) as history's greatest invention, TBWA has made a case for the biscuit as a British icon that continues to delight, surprise, and unite people across generations. I mean, can't most problems be solved with a cup of tea and a biscuit?