-
Dennis M
8I've been studying the BALO theory for the past couple weeks and wanted to share my findings:
I define a BALO within the deTERMINATOR as a magnitude 1 horse according to the AI Line, but not a magnitude 1 horse in either the AI BH, Pace E, Pace EP, Pace LP, or Pace SR.
My data ranges from 6/1/24-7/14/24 and covers the following tracks: AQU, BTP, CD, CNL, CT, DEL, ELP, EMD, EVD, FL, GG, GP, HAW, IND, LRC, LRL, MNR, MTH, PEN, PRX, PID, SA, SAR, TDN, WO
I eliminated races not handicapped by the AI and races with less than 5 entries. I also excluded races off the turf. For my research, a FTS can not be a BALO unless it was not a magnitude 1 in AI BH.
Initial findings:
2,492 BALO's with an average price of 2.9-1 and $1 WROI of .791. With an average win pool takeout around 17%, these perform worse than throwing darts. The important thing is they represent a large portion of the win pool.
Small tracks:
For whatever reasons, I have always struggled or broke even betting at what I call the C tracks. From the above list, I am referring to BTP, CT, DEL, EMD, EVD, FL, MNR, PEN, PID, TDN. I have a list of D tracks not included in my above list that I rarely play if at all. I've never fully bought the horses don't run to form as much at the smaller circuits, but maybe there's something to it. I more think there's larger issues like trainer shennanigans, lack of data integrity, tougher betting competition (less casual money). My point with this if I remove the C track from my data, here is the BALO performance:
1,356 BALO's with average price of 2.9-1 and $1 WROI of .757. That's a major difference and now we're really cooking when it come's to reducing the takeout.
Dave, any thoughts on the performance at smaller tracks and why BALO's may run better there?
Quick numbers on surface/class for the non-smaller track BALO's:
Fast Dirt: .755 $1 WROI (684 horses)
Turf: .783 $1 WROI (411 horses)
Wet Dirt: ..841 $1 WROI (38 horses)
Poly: .721 $1 WROI (228 horses)
ALW/STK: .811 $1 WROI (636 horses)
CLM: .745 $1 WROI (476 horses)
MSW: .744 $1 WROI (115 horses)
MCL: .586 $1 WROI (134 horses) -
Dave Schwartz
413
First, really good work on this.
A few comments to begin
- I am a little disappointed that they were ONLY -21% losers, but considering that these are the horses who are usually considered the best bet at the track, I can't complain much.
- I wonder if the Double BALO races might make a difference in that.
- Also, could take a look at the 1.21 and below BALOs to make it stronger.
As for your question about smaller tracks, yes, I have seen similar before. I think it ties to top trainers, because at the small tracks there are so few really "top" trainers. -
Dennis M
8
I didn't post it, but the more BALO's in the race....the more the opportunity. Double BALO's do seem like the sweet spot since there are not many races with 3+ BALO's. But instead of one over-bet horse at around 2.9-1, you have two. That's a pool you want to be in. -
Dave Schwartz
413So, the stats were actually better.
Hmm... going to mull that.
Perhaps in the relatively near future we'll be able to capture the data and take a look at grades as well.
I wonder if, for example, a BALO w/B+ performs differently than a BALO w/D. -
Jim Michalak
41dennis: great work on the study. i have a question: isn't your definition of a balo stricter than dave's? i thought a horse could be a balo if it was a 1 in aibh but not a 1 in all 4 pace categories simultaneously? i may be wrong. way to get us going, though! -
Dennis M
8
Maybe I worded it poorly, but it should be the same as Dave's. Any chink in the armor is a BALO. Any magnitude that is not a 1 is a BALO. -
Dave Schwartz
413
-
Dave Schwartz
413Do these BALO's win at about the same rate as their projected AI odds? — Slim
Since @Dennis did the study, he can provide a better answer. But as I recall, the net loss was around 25% - a far cry from the normal loss.
But I think you were asking a different question.
The simple answer is NO, they don't.
The goal of the AI LINE/BALO process is not to predict either a hit rate or the odds. It is to predict how the horses SHOULD BE BET based upon past history.
_______________
In my spare time, I'm working on the reverse of this, BTW. That is, the relationship between AI LINE and final odds, with the idea of using a pseudo GATE ODDS figure.
The concept would be to create FAKE GATE ODDS - something like half way between AI Line and final odds. Thus, a horse that was (say) 32/1 on the AI Line but was bet down to (say) 25% the FAKE GATE ODDS (i.e. 8/1) would be live.
Not suggesting that you'd actually GET 8/1 - more likely to get 3/1 or less - but such a horse has no business being 8/1, let alone 3/1. Such horses MAY turn out to be hot plays. -
Dennis M
8
I can't answer this exact question as I did not track Dave's AI Odds, but I do have my own proprietary projected odds that do a good job projecting closing odds (accounting for takeout and other horses within a given race).
My data is now updated through 7/28/24. BALO's win at exactly 25% rate. My projected odds converted to probabilities comes to 28.5% for the BALO's. Across so many races, this is a very good indication of identifying horses who will no live up their closing prices. -
Dennis M
8
My data is now updated through 7/28/24. Here is the breakdown of finishing position of all BALO's. Remember some races have multiple BALO's:
1st - 665 (25%)
2nd - 517 (19.5%)
3rd- 414 (15.5%)
4th - 331 (12.5%)
x (5th or worse) - 728 (27.5%)
Total - 2,655 horses
As I said in another reply, these horses command a 28% expected win rate according to my proprietary projected odds metric. -
Dave Schwartz
413I was just thinking... since all favs win at ~39%, and these BALOs are (likely) to be lower odds than the average favorite, that would put the IV probably around 0.65 or so.
-
Dennis M
8
I find that if a BALO is present, it tends to be a more chaotic race vs. a less chaotic race. Larger fields, more chaotic conditions, tepid favorites in general, etc. In some ways, it is also useful as a race selection filter. -
Dave Schwartz
413I find that if a BALO is present, it tends to be a more chaotic race vs. a less chaotic race. Larger fields, more chaotic conditions, tepid favorites in general, etc. In some ways, it is also useful as a race selection filter. — Dennis
YES!
That was one of the points I originally made!
Recall the talk that begins with... "You have a horse who is 3/5. He got that way because he really is the best horse in the race. (when you look at his factors.)"
"If the handicapping works, this horse is the obvious winner."
"Therefore, one could say that for you to cash a ticket on another horse, THE HANDICAPPING MUST FAIL!"
That's the very definition of chaos! -
Dave R
28Great discussion guys! Man, there is very little doubt in my mind that there are different levels of BALO's and that we will be able to improve our level of success in them. For instance, I have noticed that if a horse may have a rank of 2 and a margin of 2 on only 1 of the pace lines it is most likely a marginal BALO at best. If it doesn't rank at all on several lines, it is a strong BALO. I am beginning to think that we need a tool to calibrate the AI so that we may make it stronger or at least make our elimination processes stronger. -
Dave Schwartz
413For instance, I have noticed that if a horse may have a rank of 2 and a margin of 2 on only 1 of the pace lines it is most likely a marginal BALO at best. If it doesn't rank at all on several lines, it is a strong BALO. I am beginning to think that we need a tool to calibrate the AI so that we may make it stronger or at least make our elimination processes stronger. — Dave R
Great idea.
I will add this to the tracking process.
I like the idea of BALO1 & BALO2, based upon what Mags they were eliminated on. -
Dennis M
8Hey guys,
6 months later and long overdue, I wanted to give an update on my BALO analysis. Over time I was able to test "chink in the armor" BALO's from 11/1/23 to 9/22/24. I was able to track 10,560 BALO's during that period. At some point, I gave up on small tracks (diminishing returns) and focused only on the following tracks: AQU, CD, CNL, DMR, ELP, FG, GG, GP, HAW, IND, KD, KEE, LRC, LRL, MED, MTH, OP, PIM, PRX, SA, SAR, TAM, TP, WO.
Of the 10,560 BALO's, they return $0.776 per $1 dollar wager. These BALO's go off at odds of 3-1 on average.
As a reminder, the definition of a BALO is magnitude 1 horse according to the AI Line, but not a magnitude 1 horse in either the AI BH, Pace E, Pace EP, Pace LP, or Pace SR.
Personally, I use some of my own proprietary data to filter these BALO's to where they return about $0.74 per $1 dollar wagered. I'm sure there's other ways w/in DET for you to find the best BALO's to fade.
My question to the group is what other BALO methods are being used by others? I might have the data accessible to test further BALO theories.
Dennis -
Jim Pommier
127Wow-- 10,560. Do you have the win percentage for the BALO's? Also, "AI Line, but not a magnitude 1 horse in either the AI BH, Pace E, Pace EP, Pace LP, or Pace SR"-- did you track the win percentage for say BALO's that failed more than one of the 5 tests? For example, BALO's that failed AI BH plus 2 of the pace tests? Curious if more fails lowered the win percentage. -
Dennis M
8
Their win rate is 22%. I collected all of this manually, so I sacrificed depth for breadth. As described, I only looked at if they failed to have a magnitude of 1 in the AI BH, Pace E, Pace EP, Pace LP, or Pace SR. -
Jim Pommier
127Checking to see if we can leverage your work. You may have saw the other thread in reference to starting a BALO research study. I spoke to Duffy and we were thinking to start with 1,000 of your BALO's-- Date, Track, Race# and Horse#. Duffy said he could then bang those races in a study to test against the database and share the results with the group.
-
Dennis M
8
I'm not sure what you're asking for. Didn't I just share the results with you? Or am i confused? -
Jim Pommier
127Sorry, we're looking to see if you kept the actual data-- Date, Track, Race# and Horse #.
-
Dennis M
8
-
Jim Pommier
127Duffy (PickMaster) was going to test the BALO's using over ten years of data in his HSH database (the Model folder). We're hoping to further identify losing BALO's.
Ping:
@PickMaster
Please register to see more
Forum Members always see the latest updates and news first. Sign up today.