How a Professional Punter Rebuilds a Model That Has Stopped Working

Recorded August 2025 for The Grind Ep 4. Any figures mentioned are from that time.

When the numbers turn against me, I do not throw the model out and start again. I go back through it piece by piece. A lot of this work I do whether I am winning or losing, but I do it more closely when I am losing. The process is the same either way. Check the factors, rerun them, look for small improvements, then check that each part of the system is carrying the right amount of weight in the final figure.

Going back through the factors

The first thing I do is go back and check through all the factors. I rerun a lot of them. I look at whether I can improve them. I add in different variables to try and get the bottom line a bit better.

That might mean bringing the pace of the race into a factor that did not have it before. It might mean excluding really high paced races out of a factor altogether, then reducing that factor by something like twenty per cent to improve the model.

None of this is dramatic. I am looking for half a per cent here, one per cent there. These are not big swings. They are small changes, and over time small changes add up to big changes. If you go in expecting to find one broken thing that explains a losing run, you will usually be disappointed. It is rarely one thing.

The system is many models, not one

The other part of the work is reviewing each model on its own. My system is broken down into many different models rather than one big one.

There is a model for mapping. There is a model for speed ratings. There is a model for sectional ratings. There is a model for video comments. There is a model for jockeys, and there is a model for trainers. All of those models come together to form the final numbers that go out through the website.

Breaking it up like that matters. If everything sat in one block, I would have no way of telling which part was dragging and which part was doing the work. Separate models let me look at each one in isolation and ask whether it still stands up.

Checking validity and weightings

So what I am doing is checking two things on each model. First, the validity. Does the model still hold up when I rerun it. Second, the weighting. What is that model contributing to the overall result, and have I got that contribution right.

Sometimes I find I am over factoring something too heavily. Sometimes I find I am not factoring something else in enough. When that happens, I change the weightings slightly. Not a wholesale rebuild, just an adjustment, then I look at the overall picture again and see where I sit.

That is the point of the exercise. I want a clear view of exactly where I am and what I need to be doing to keep moving forward.

If you build your own numbers, the takeaway is to split them into parts you can test separately, then go hunting for the small gains rather than the one big fix. Half a per cent at a time is slow, but that is how the work actually gets done.

Every race on the card, rated, with Kingsley's selections and the speed maps behind them. Your first day is free, no card needed.

More from Kingsley

• "Why Is It So Hard to Price a Race From Ratings Alone?"
• What a Time Rating Is and How Class, Speed and Weight Go Into One Number
• How I Compare Top Mares on Peak Ratings Over the Last 600 and 200

From the full video: The Grind Ep 4 | Turning Ratings Into Profits – Real Betting Analysis & Mindset in Action

Gamble responsibly. 18+. If gambling is a problem for you, call Gambling Help on 1800 858 858 or visit gamblinghelponline.org.au