tony spencer
Colt
Trainer Race Intelligence – What the Dashboard Is Doing
I’ve been developing the Trainer Race Intelligence page a bit further, and it is now doing quite a lot more than simply ranking trainers.The idea is to bring together three separate strands of evidence and look for occasions where they agree:
- Trainer Modular model – looks at the individual runner and whether it fits the trainer patterns that have historically been productive or negative in similar circumstances. This gives the trainer/model rank and percentage within each race.
- Market / BFSP model – uses the Bet365 price journey from the previous evening through 8am and 10am to estimate where the horse is likely to start on Betfair. It shows both a normal Predicted BFSP and a Smart BFSP after allowing for trainer behaviour and the strength of the move relative to the rest of the race.
- Trainer-money behaviour – looks historically at whether money for that particular trainer has actually been worth respecting. Some trainers have genuinely strong results when their horses are backed; with others the money has historically been misleading. The same work is done on horses that drift.
For example, a horse can show what looks like very shrewd money, but if the Trainer Modular model ranks it 10th of 10 in today's race, the two systems are disagreeing. That horse is now shown as a SHREWD MONEY WATCH, rather than being promoted to one of the day's best bets.
The strongest selections are therefore based on convergence.
A STRONG or VERY STRONG BACK is where the trainer/model evidence is positive and the market evidence is also supportive.
A STRONG or VERY STRONG LAY needs an existing negative trainer/lay profile which is then confirmed by things such as a significant drift, False Money behaviour or a historically poor trainer-money pattern.
This makes the headline selections much more selective. Interesting market moves which don't have enough trainer confirmation are still shown, but they sit separately under Money Watches or Drift / False Money Watches rather than being presented as bets.
There is also a separate Grade B & C section based on the frozen trainer research. These identify trainers whose historical results have been particularly interesting when their runner falls within certain positions in the Trainer Modular rankings.
For every runner you can see the evening price, 8am price, 10am price, Trainer Modular rank, model %, predicted BFSP, Smart BFSP, expected market move and the current signal.
Clicking on a horse takes you directly into that race, and clicking the trainer opens the underlying Trainer Behaviour information, including sample size, historical win rate, BFSP expected rate, A/E, POT and whether the evidence is emerging, strong or established.
The important thing is that the page is not simply saying “this horse has shortened, therefore back it”. It is trying to answer a more useful question:
Does today's runner fit a historically strong trainer pattern, and is today's market behaviour confirming it?
When both independent strands point the same way, that is where I am most interested.
It is still very much a research tool rather than a tipping service, but it is getting closer to giving a quick view of where the strongest agreement — and strongest disagreement — exists on the day's card.
Live dashboard:
https://trainer-race-intelligence.netlify.app
