Co-Founder @ SYNTAX | Applied AI Engineer | Building & Scaling AI Products, Sports Prediction Engines, SaaS Platforms & Agentic Systems | Built Gamelens
Want to connect with founders, CTOs and product leads in sports tech: prediction, analytics, fantasy, fan engagement.
I build AI products for sports: ML models, real-time data pipelines, AI analysts and native apps.
Building in this space? Say hi ๐
@hockey_gamebot the win probability swing on each goal is a great way to tell the story of a game. are those swings worked out live during the game, or only once the tracking data lands afterwards?
@GreatTennisBets 5-0-1 off the model in a week is a good run. is it built mostly on serve and return numbers, or are you adjusting for surface and fatigue as well?
@fpl__core@fplcopilot@SpirosValouxis pulling Copilot's xP into the team planner is a smart tie-up. are you fetching it from them on a schedule, or do they push fresh numbers to you after team news drops?
@Karan_Talwar27 Miller at 16% from three on 13 attempts a game can't hold. how many games do you give a shooting slump before the spreadsheet starts believing it?
@_HonestDegen a 78-source readout before tip is a lot to pull together for one player. are you scraping most of those yourself each night, or does a lot of it come in from a feed?
@UDLRsports Reyes landing that KO after the model flagged him is a good advert for it. does the UFC model weight recent fights more heavily, or treat the whole record about the same?
@Roboteach7@InsidersEdge7 Monte Carlo for goal scorers is a fun way to use it, and 3 of 6 is a good night. is the engine mostly shot rates per player, or does power play time and linemates feed in too?
@Dior_Bets a million sims a game is a serious setup for a full slate. are the injury updates going in automatically before tip, or are you fixing rosters by hand each night?
@GMGoBetting grading it against the number at puck drop is the honest way to do it. is the model mostly shot rates and ice time, or are you adjusting for score effects and the opponent too?
@fplcopilot Saka out for Dewsbury-Hall is a move I didn't see coming. how far ahead does Copilot plan transfers, just this gameweek or a few weeks out with the chips in mind?
@EngineFant an AI assistant sitting on top of your own model is a nice idea. is it reading straight from the R projections, or from a summary you hand it each gameweek?
@FPL_Market_Edge 81.6 projected minutes to one decimal is a nice touch. do you model minutes on their own first and multiply through, or does it all come out of the one model?
@ArmChairLogan 12 hits and now no Goedeke is a rough combo for a first game back. does your model knock the line down for a backup tackle directly, or does it only show up once the pressure rate moves?
@WalrusQuant the key numbers part is where plain poisson really falls over for nfl margins. did negative binomial handle the 3 and 7 spikes, or did you end up layering something on top for those?
@NukesOnlyVIP@IntelliPickz makes sense, usage tells you a lot more than the raw ice time. do the last few games get extra weight, or is the whole recent stretch treated the same?
@SAAdvantage makes sense, flagging it internally is the right place for it. sounds like the other workaround is part of the secret sauce, so I won't push ๐ looking forward to the team comparisons
@LockItDownSport@NHL@Senators going through all 32 teams one at a time is a big project. are the player point totals coming straight from your own model, or adjusted by hand from last season's numbers?
@FPLDaveHQ the six gameweek fixture view is always the bit I check first. does your points model feed straight into which fixtures you target, or do you run those two separately?