Trading Out at Profit vs. Letting the Bet Run – Long-Term Edge

[globi166]

If I understand you correctly, you're referring to odds in general without applying any selection criteria. However, if you use a selection process—for example, there's a video and template on this forum showing how to spot market bias—then things can look quite different.

One example is identifying a trainer and jockey combination that has already won 2 or 3 races on the card. This often leads people to start backing another runner from the same combo quite heavily, which can distort that horse’s price. If we focus only on this type of market movement, the trading picture you posted would look very different and potentially show clear value opportunities.
Or I'm missing something?

[ShaunWhite]

It was just a response James saying the markets are efficient, and you questioning that because you'd seen a 1.08 horse lose. I thought it was a reasonably interesting explanation of it

But as you say the true price (efficiency) is only evident over large samples and each individual one is undoubtedly wrong. Not just because it would be statistically impossible for them all to be right, but it's also the case that the true price of any horse on any given day (to within a meaningful margin) is unknowable. They're horses not dice.

The things you mention about jocks/trainer combos winning and causing subsequent markets to be affected is fair enough, and that's something that happens in all markets everywhere. Humans over-react to news. Then when they assess it more calmly they revert to some sort of sense. Happens in politics with polls, football with goals and F1 with quali, racing with trainer wins. And the moves that causes means you can get structural EV (moves) without necessarily knowing or caring what the 'true price' is. But having a decent idea of true price stops you giving away the EV you've earned on the move by having sub-optimal bets at each end of it. The structural EV you get is just the sum of individual EVs you got while creating that position.

If I've missed the point I've wasted my time there, but nevermind.

[ShaunWhite]

It's a good jobs it's complicated, if it was easy everyone would get it, and nobody would make any money

[globi166]

It was just a response James saying the markets are efficient, and you questioning that because you'd seen a 1.08 horse lose. I thought it was a reasonably interesting explanation of it

But as you say the true price (efficiency) is only evident over large samples and each individual one is undoubtedly wrong. Not just because it would be statistically impossible for them all to be right, but it's also the case that the true price of any horse on any given day (to within a meaningful margin) is unknowable. They're horses not dice.

The things you mention about jocks/trainer combos winning and causing subsequent markets to be affected is fair enough, and that's something that happens in all markets everywhere. Humans over-react to news. Then when they assess it more calmly they revert to some sort of sense. Happens in politics with polls, football with goals and F1 with quali, racing with trainer wins. And the moves that causes means you can get structural EV (moves) without necessarily knowing or caring what the 'true price' is. But having a decent idea of true price stops you giving away the EV you've earned on the move by having sub-optimal bets at each end of it. The structural EV you get is just the sum of individual EVs you got while creating that position.

If I've missed the point I've wasted my time there, but nevermind.

Yeah, that makes a lot of sense — appreciate you taking the time to break it down. I totally agree that individual prices are always wrong to some degree, just because of how unpredictable racing is — like you said, they’re horses, not dice.

The idea of structural EV is really interesting. I hadn’t thought of it that way, but it explains well how short-term overreactions (like trainer/jockey hot streaks) can create temporary edges, even if we can’t pin down a precise true price. I guess the key is having a decent enough idea of value so that we’re not giving it back by entering/exiting at the wrong points.

Also agree on market behaviour — you see it everywhere, not just racing. People chase narratives, overreact to recent outcomes, and then the market eventually settles. And I suppose if you can catch that imbalance early, that’s where the edge lies — not necessarily in knowing the future, but in knowing when others are mispricing it.

Anyway, great explanation — thanks again.

[ShaunWhite]

Anyway, great explanation — thanks again.

Any excuse to talk about it. Took me a while to see that where red and green meet on the ladder is the average price you've got after all your backing and laying, ie the EV you've built. And if you've got a green at the price that's likely to be sp, you soon start to ask yourself if you'd have been better off 'trading' smaller amounts and just letting them run. Afterall beating sp is the betting Holy Grail. But.... SP is 0ev (or is long term) so if you hedge at 0ve it neither adds to your value or reduces it, those bets are a wash. Hedging reduces your commission overall and your variance, but betting-wise (gross ev) it makes no real difference.

From an automated perspective, where bets are often placed on perceived current value (ev) rather than the prospect of a move (structural ev) the hedge bet is treated like any other bet. It's either a value bet or not using the same algo as for the preceding bets. Long term performance is more important than equalising your position, so you always stake such that the hedge is optional rather than a matter of life and death with a grand on the table. And in-running especially, staking too big carries the risk that in order to get out you have to give away some of the value you've earned. It was additional risk that actually produced a smaller return.

It's definately a 3 pint subject rather than 128 characters.

[globi166]

It's definately a 3 pint subject rather than 128 characters.

I already struggle to get my head around it sober — imagine trying to make sense of it after a pint! I'd need subtitles and a whiteboard lol

[sionascaig]

From a profit maximising perspective you should treat every trade individually on it's own merits and trade in or out only where you are expecting a positive expected value (+EV) from doing such. If you have +EV from your original trade, but no +EV from your exit then you will be giving back some of your profit in the long term. Consider it as a cost for reduced exposure and better bank control. Is it worth that cost to you?

Currently messing about with a dog bot that is value centric & still early days but seeing:

Profit on Backs + Profit on Lays (by selection) is greater than Profit on Backs or Profit on Lays.

In other words the EV on (Lay + Offset) > Max( EV on Lay, EV on Back)

==> that is not what I was expecting...

(note: not every lay has an offset matched)

Presumably this is a corollary to what you are getting at : it is possible to exit for an overall greater profit provided the exit point is +ve EV AND is greater than the bit of EV you are giving up by exiting?.

[ShaunWhite]

What are you using to measure Ev? Bsp?
It's perfectly possible to back above and lay below sp in one market, so hedging doesn't necessarily mean a zero or a loss on the bet that equalise your position.

[sionascaig]

What are you using to measure Ev? Bsp?
It's perfectly possible to back above and lay below sp in one market, so hedging doesn't necessarily mean a zero or a loss on the bet that equalise your position.

I should have said "profit" rather than "EV" as it is just the profit I'm currently looking at (nothing "expected" about it)...

Only just got a few days of data so it might just be some freak variance..

e.g. some of the lay's could be potentially -ve EV (tbc) and the backs have created a +ve that otherwise would not be there - really not sure yet...

An example from yesterday follows (all the backs are offset bets & occur c10% of the time).

Screenshot 2025-06-30 160416.png

[ShaunWhite]

I tend not to look at cash esp over the short term, just the EV. With enough races you'll find that the pl line for cash and the pl line for ev track each other and converge. If you end up dead level across the card there's no difference but if not the actual wins and loses make a big diffence.

But breaking down bets like you are definitely helps, just for your own interest, if you've got the EV per bet (price vs the sp) then sort them by price and do a pl chart for that sequence. Cumulative PL vertical, price horizontal. You might just see a down slope and an up slope (not random wiggles) and it might point to price bands you're more effective on. And with pl (ev) per bet you can also sort them by things like time between bet and scheduled start (countdown) , early or late can be good or bad for you. Your bet history is the info you buy when you try something out.

Be careful not to overfit though, the good and the bad can easily revert and swap places, and you end up chasing the most recent setup. That almost guarentees losses because you bet things that used to work just in time for them not working again.

[sionascaig]

Thanks Shaun, that is very helpful...

Will continue & gather data for now (on piddly wee stakes) until I have something to crunch )