AIPaddock.com

Want to take a more evidence-based approach to @photofinishgame? Then bookmark this thread of bite sized articles exploring different aspects of the game. 16 blog posts with new angles and alpha. Thread: 1/17.

Updated Righties for season end.

breaking news, that data didn't include the last weekend of the season where about 50% of the records were broken. sorry. corrected figures below. Pappy On the Point broke 2 records last weekend (szn32 2yo only).

Fasted 2yos from szn32 of @photofinishgame. grab a microscope. Lefties first. All combos included. Loving the audacity of Tutankhamun I.

@Poor2Ricopfl @cadetracing @dsilver88 @photofinishgame my guess is that there is a mathematical peak for every horse, but for some that might cover 2 distances. this is based off of their start, speed, stamina and finish attribs. luck/noise distorts the picture though especially on low run count.

@ai_paddock @cadetracing @dsilver88 @photofinishgame From your stats. Do you see most horses performing better at 1 dp....2 dp....or 3 more distance preferences?....just trying to see if the average horse is mostly good at...without looking at the stats i say most excell at 2 distance preferences but wanted ur educated opinion

Some new contender 2yos in week 2.5 of season 32 @photofinishgame Trueskill ratings, or were they just lucky times? (#2 ;p)

each horse has a distance or two that it performs best at. either side of those distances, performance slopes off. so there is some adjacency relationship. there is a bunch of randomness applied so it is often a murky picture, but looking at the FF chart and comparing peaks or troughs often gives you a straight line relationship.

@ai_paddock @dsilver88 @photofinishgame As far as distance is there a bump if you are really bad at 4f/5f to imply you are really good at 11f/12f and vice versa?

@SecretariYat @dsilver88 @cadetracing @photofinishgame yes - its laziness. quick fix - just show SS-/S+. long fix, appraise different races differently...

@ai_paddock @dsilver88 @cadetracing @photofinishgame Wouldn’t it make some sense to not combine horses running in grade restricted races with those running in open races?

Trueskill is a Microsoft algo, designed for pvp games to put equal players in the same lobbies. PFL is different because each distance is like its own game. To do what you are suggesting is to create your own handicap model, which is what I would guess any gambler has done. Custom models should be 3-5x more predictive than ML. Trueskill just looks at finish position and the difficulty of the competition - hence low grade horses can be scored high if they beat other low grades and never meet higher grades. It’s a pure pvp metric

@ai_paddock @cadetracing @photofinishgame Have you considered including odds into the ratings? (if they aren't yet)

@cadetracing @dsilver88 @photofinishgame Yeah the algorithm has some quirks, but it just tries to separate skill from luck. getting best distance early skews the results, as does early age form. I see Tempus Fortuna and Maid in the Grass just won their big 2yo races so sometimes it can be useful to people.

@dsilver88 @ai_paddock @photofinishgame And what the heck do you know? How majors have you won this week?

Just placed a $180.00 Pick3 bet on Twin Spires Sprint! Tail me? 🎯 Picks: 3,6,8 / 10,15,17,18 / 7,8,17 @betvibewin 🔗 beta.betvibe.win/races/season00…

Big race day in Season 32 of @photofinishgame needs a scatterplot. Last season a great horse with average form won. Today's conditions suit the LDS horses.

Just placed a $180.00 Pick3 bet on Artemis Cup Sprint! Tail me? 🎯 Picks: 5,16,17 / 2,4,6,17 / 1,12,13 @betvibewin 🔗 beta.betvibe.win/races/season00…

@Gavone72 @photofinishgame Fingers crossed!

Thanks for the free breeding reports @photofinishgame Just ran one for my Scott Pendlebury - Big Wick Energy combo... confirming that if I am not super unlucky (which I normally am - who can forget my run of boys early in the game!?!?!), I will be in the SS Club in a week!!

Just placed a $120.00 Trifecta bet on GRADED Road to the Twin Spires Sprint R77! Tail me? 🎯 Picks: 4, 10, 11, 13 (Boxed) @betvibewin #betvibe 🔗 beta.betvibe.win/races/season00…

Just placed a $120.00 Trifecta bet on GRADED Road to the Twin Spires Sprint R77! Tail me? 🎯 Picks: 4,10,11,13 / 4,10,11,13 / 4,10,11,13 @betvibewin #betvibe 🔗 beta.betvibe.win/races/season00…

@VintageTent @photofinishgame that can clearly be a thing, but because its win rate, hopefully there is still some signal in the fact that horses that run off preference do badly (as you would expect)

@ai_paddock @photofinishgame Interesting. I wonder if that is just self selection bias or not

How do conditions preferences work in @photofinishgame? a. linearly b. weirdly (check out 0 fast stars performing worst across all types on fast tracks). Also 3 fast stars become a big hinderance on sloppy tracks. Finally yielding is not a mid-point, its a different multi-verse where 0 stars live and thrive (fast and slow don't get say half their usual bonus - stars are bad at yielding).

@stables_guy @photofinishgame there could be some other factor skewing stuff, but generally there is no relationship between attributes and stars. the sample here is pretty much all races ever. its probably fairer to say yielding is neutral to stars.

@ai_paddock @photofinishgame how do you know attributes or some other factor is not skewering results. the fact 0 star dominates on yielding seems off. if 2 horses have 3 stars direction and surface, but one has 3 stars firm and other has 0 stars firm why would 0 stars firm be better on yielding

@WuDavide7 @photofinishgame sorry I am mixing up terminology here a bit. Fast and Slow means Firm and Soft as in RDF vs RDS

@ai_paddock @photofinishgame Sorry what is the meaning of fast vs slow?