Finding the Poppers

[Ray2000]

I'm looking at 217,324 starters in 2014 where 25,930 of them hit the board at Odds > 8/1
I wanted to see what were the most important factors that can be as flags to include them in exacta/trifecta boxes,
even if they don't figure to be contenders by conventional 'capping.


Rankings are by ChiSquare p values

5E-20 Class Down, Trackmaster 4 or more drop from previous race
8E-12 Class Up
3E-07 Trainer Change good or bad
0.002 Lasix Change
0.006 Different Track than last race
0.011 Driver Change positive (+50 UDR)
0.019 Post Draw (4 positions or more to the inside from last race)
..............Any p value > .05 is considered insignificant such as
0.172 Driver Change any
0.210 Driver Change negative
0.630 5 hole draw


Any ideas on what other flags I should be looking at.

[lamboguy]

distance changes, length of time between races, weather conditions (temperature, precipitation and wind)

[RaceTrackDaddy]

Ray:
I would look for first start in a new barn, ownership (with same trainer or new trainer change) with a sub program listing the results with respect to trainer percentage (0.200, 0.250 and 0.300 and above).

rtd

[Ray2000]

Lambo

Distance changes and weather conditions are hard to do. There's just too few races with distance changes to get a good number and I don't have wind and such.

As for days off, the average was 8.45 for the longshots compared to 8.65 for the others and would be considered statistically significant. (pvalue 2E-05)

A good example of the weather helping a 50/1 shot get 2nd was Northfield R13 last night. (below) The 3 horse found himself in the garden spot when the 1 stumbled at the gate pushing the 9 and most of the outsider back enough that closers were out of it in that blizzard.

It's most likely that there is No "tell" when trying to find these poppers, but I'll keep looking.:)

Bobby

Going from a dud to a star on the barn/trainer angle is highly important but I couldn't find enough longshots with that factor to be statistically meaningful. The crowd just won't let them go off that long.

Thanks for input.

[pandy]

I'm looking at 217,324 starters in 2014 where 25,930 of them hit the board at Odds > 8/1
I wanted to see what were the most important factors that can be as flags to include them in exacta/trifecta boxes,
even if they don't figure to be contenders by conventional 'capping.


Rankings are by ChiSquare p values

5E-20 Class Down, Trackmaster 4 or more drop from previous race
8E-12 Class Up
3E-07 Trainer Change good or bad
0.002 Lasix Change
0.006 Different Track than last race
0.011 Driver Change positive (+50 UDR)
0.019 Post Draw (4 positions or more to the inside from last race)
..............Any p value > .05 is considered insignificant such as
0.172 Driver Change any
0.210 Driver Change negative
0.630 5 hole draw


Any ideas on what other flags I should be looking at.


Good idea. I think the post position factor could be a key. Just this month I started a test on my trotpicks.com website where I am going back and analyzing all longshot winners at 14-1 or higher at the tracks I follow, Balmoral/Maywood, The Meadows, Dover, Monticello, Yonkers, Meadowlands, Cal Expo, Northfield, Freehold. Just to see if there are any factors that keep repeating.

[Ray2000]

I agree and the post draw improvement compared to last start (+4) that I posted, does need more work. It should be track specific and use best/worst posts by winning percentages rather than just drawing 4 positions "better" or inside.

Other track specific factors might be.. Morning line alone, Morning line - Final line ratio, and some toteboard "tells" such as Win pool odds compared to win odds by Exacta pool action.

GL

[pandy]

Another good longshot angle may be second or third start off a layoff.

I also wonder if a negative driver change is a legitimate angle. Sometimes a horse is bet with a top driver and loses, then the next start it has a driver who does not have a high win percentage but it wins and pays big.

[traynor]

Long odds are less a factor of the horse that won as they are a factor of the horse(s) that were heavily bet and did NOT win. That is, it may be more useful to consider what else went on in the races you are studying, other than (only) attributes of the winner.

For example, some drivers at some tracks (Yonkers is a good example) seem to be both overbet and tend to finish up the track at very low odds. When you locate a number of such factors and consider them in your calculations, a far different picture emerges.

You might consider the attributes of a horse that won at generous odds if-and-only-if (for example) another horse in the race went off at 2/5 or less. Similarly, consider the attributes of a horse that won at generous odds if-and-only-if no other horse in the race went off at 8/5 or less (the so-called "chaos" races). The key element is not so much finding the attributes of the long odds winners as it is finding which underlays can be safely (or profitably) tossed, or which other race factors contribute to the longer odds entries winning.

[mrroyboy]

The best use of this is to eliminate losers.

[LottaKash]

IOther track specific factors might be.. Morning line alone, Morning line - Final line ratio, and some toteboard "tells" such as Win pool odds compared to win odds by Exacta pool action.

GL

Ray, especially now, isolating and using with emphasis, on the Trackmaster ML tracks..

[mrroyboy]

I agreed John but how would you recommend we use the Trackmaster ML?

[Ray2000]

All good points guys, and this factor search is a bit (maybe very?:)) confusing because I'm not looking for longshot winners that are good overlay bets* but rather "What do non-contenders have in common that surprises the hell out of everyone when they hit-the-board at big odds?" And you might say "Hell, that's just another way of looking for longshots" or "What are you calling a non-contender? or "Just dumb luck"

And I get that... but I still think there might be a combination of attributes on a non-contender that indicates this is the type of horse you don't want to leave out on the exotic ticket combo.

So to begin, I'm just looking at the frequency comparisons on different factors. Examples

How many gelding starters are there in the longshot winner group (as a percentage), vs gelding starters overall.
Answer is...Statistically insignificant.

How many class-up starters are there in the longshot winner group vs class-up starters overall.
Answer is...Statistically very significant. Pandy's comment on "negative driver change" fits in here well as this is usually considered a 'negative' class change.

I'll update this if any strange results are found.

And yes, LK, that means building a new data base for the tracks using Trackmaster ML. :rolleyes:



*That would have to take into consideration Traynor's point on examining the competition in a specific race.

[mrroyboy]

So moving up in class is a good factor for picking longshots?

[RaceTrackDaddy]

Thanks Ray

[Ray2000]

So moving up in class is a good factor for picking longshots?
No I can't say that, roy.

At this point all I know is ...
In a group of 25,930 longshots who hit the board ... 14.6% were Class-ups

compared to the rest of the starters (191,394) where 13.1% were Class-ups

The question to answer is...Is this 13.1% vs 14.6% statistically meaningful?

The math says it is extremely unlikely that this would occur by chance so Class-Up is a factor to consider.

[traynor]

No I can't say that, roy.

At this point all I know is ...
In a group of 25,930 longshots who hit the board ... 14.6% were Class-ups

compared to the rest of the starters (191,394) where 13.1% were Class-ups

The question to answer is...Is this 13.1% vs 14.6% statistically meaningful?

The math says it is extremely unlikely that this would occur by chance so Class-Up is a factor to consider.

That is where the answers depend on the questions asked. Up in class from last race is WAY different (usually negative ROI) than up in class from second race back--which can often be used as a standalone positive ROI factor (varying by track).

The "second start after whatever" phenomena is pervasive, and you might find it useful to include in your analysis as a separate and distinct category.

[Lose The Juice]

Longshot winners these days do not tend to have across-the-board commonalities, because different trainers win off different patterns, many of them not detectable from the program lines in any given race.

Hint: if you're missing last-out scratches for whatever reason, you're probably overlooking a lot.

[Ray2000]

1. "second start after whatever" phenomena

2. "last-out scratches for whatever reason"

A couple of interesting leads, I'll check them out as soon as I get time.
thx

[mrroyboy]

I like the second time anything. Logical. First time didn't work. Now we know the horse better. Second time works???

[Lose The Juice]

"Up in class" can be very subjective, especially in these days of conditioned racing. If the race below Preferred is usually written as NW28 L5 but can't fill at that level, the racing sec may write it as a NW 30 or 32, albeit with most if not all of the same horses.

[Ray2000]

"Up in class" can be very subjective, especially in these days of conditioned racing. If the race below Preferred is usually written as NW28 L5 but can't fill at that level, the racing sec may write it as a NW 30 or 32, albeit with most if not all of the same horses.

Very true, that's why I went with Trackmaster's Class scale rating (60-110) for a race in the OP.
It's speed based and even though I don't particularly care for their speed ratings, an average of all speeds for all entries does give a class number that can be used for comparison of races. IMO

[Lose The Juice]

Don't kid yourself, I realize you probably need a shortcut, but there are better ones that are more based on the realities of each track. A bad metric on the class issue will skew your numbers seriously.