Building a Systematic Handicapping Framework

1. WHY WINNING IS STILL HARD

For certain surface-class combinations, you only need to consider a few simple parameters to consistently predict which horses are genuine winning contenders and which horses will only get close (ideally including the public favorites).

As I mentioned in another article, selecting the right surface and class combination is only the first step to improving your strike rate. Most punters have probably said something similar to what I’ve said:

Damn it! I knew it would be one of those horses!

Here’s the problem: if the winning horse falls within our expected range, why can’t we still make money? My intuition tells me the issue lies in how we define “reasonable range”—we lack a consistent standard. This is where emotionless AI algorithms can help me repeatedly validate whether the same horse selection method can consistently predict race outcomes.

2.1 DEFINE 3-TIER SYSTEM

My algorithm uses “beaten lengths” as the primary parameter, automatically calculating advantage scores against other horses, then ranking them into three tiers: Strong 、Mid 、Weak. My hypothesis is:

Strong
Horses that have run close to Strong horses in previous races can be upgraded. However, Strong horses themselves aren't necessarily more likely to hit the trifecta, because the algorithm doesn't account for other negative factors (e.g., rating plateaued, faced tough competition last start but overall performance was off).
Runners raced with Mid horses 
Horses that have previously lost to Mid horses should be downgraded. The concept is that Mid horses represent the threshold for this race—if you can't beat a Mid horse, you'll struggle to challenge other competitors in the field.
Runners raced with Weak Horses
Horses that have previously beaten Weak horses can be upgraded. The concept is that when these horses face Weak horses again, they have another opponent they can confidently beat, boosting their competitive drive.
Mid / Weak Horses 
If you can establish all three tiers (Strong/Mid/Weak), horses that are both Mid AND Weak should also be downgraded.

2.2 HOW TO FORM BETSLIPS

Sounds complicated? In practice, you can:

  1. First, identify Mid-Weak horses and Mid-related horses to establish a range, betting that these horses cannot both finish in the top two.
  2. Among the remaining horses, identify Strong/Weak-related horses.
  3. Third and most difficult step: Ensure your betting structure accurately reflects your final selections.

Using Race 2023-682 as an example:

  • Since the Tier table has 3 levels, the Mid-Weak horses themselves (namely: YOUNG HORIZON, LUCKY TOUCH, SUPER UNICON, GOLDERN DARCI) and Mid-related horses (SUPER LEGENDS) —these 5 horses—are determined unable to finish in the top 2 simultaneously.
  • Among the remaining horses, STRATHPEFFER is both a Strong AND Weak-related horse, giving it the highest winning probability.
  • For betting, the simplest approach would naturally be to back Dragon City Champion to Win (with 17-1 odds, no less!). But analysis often tells us the reasonable range won’t be just one horse. For example, in practice you might consider FOOLISH HEART as another contender, so I prefer using keying strategies in Trifectas/Superfectas to increase my margin for error.
  • You can even get creative by selecting an “easy to place, hard to win” horse from the Mid-Weak group and betting it for a specific finishing position (2nd/3rd/4th) in the exotic pools.

As for the specific details of my algorithm, I can’t reveal them here. After all, the essence of the pari-mutuel wagering game is pool-splitting—I can’t simultaneously hope my method succeeds while having too many people use the same method to select winning combinations, which would lower my returns. But I strongly encourage everyone to use AI tools to build your own racing theories: I highly recommend julius.ai — it can help you accurately analyze large amounts of data without any coding knowledge required.

3. WHAT’S NEXT

Horse racing is just my hobby. Until next year, I’ll only continue posting case study analyses on an irregular basis. If I can identify two or three distances with higher strike rates, I’ll provide occasional race day analysis starting next year. The principle is simple: I’ll back my selections with real money to hold myself accountable.