Hedging an Edge

[rinconpaul]

If you perceived that say horses using a certain barrier #? were profitable to lay or back long term, but it was a bit of a roller coaster getting there, how could you apply a biased hedge to it? This would protect your downside.

You're hedging a long term edge based on a material condition, not form or price differential. It might be heavy tracks v good tracks, a particular distance v another, a certain jockey, trainer and so on. Another example might be laying the fave on heavy tracks is more profitable than same on good tracks. The obvious answer is just don't lay faves on good tracks. Using options in fin markets, you can construct all sorts of counter intuitive plays.

It's not about obtaining a lay price lower than back, it's more global. Come back to my starting example. I want to lay every barrier 4 in every race. There's going to be some losing days, maybe weeks but long term it should be profitable. How can I limit or spread the risk?

[marko236]

The market its self shows where the profit is, if say a horse has traded at odds of 3 then goes down to 2.5 and at the start off the race is trading back at 3 you have to say that any lay price below 3 should be a long term edge, on the risk side if betting using 1% of your bank should give plenty of space for long losing runs, if you have an edge your bank will jump up and down but it will slowly increase over time.

[marksmeets302]

I'd say there are two answers:

One is quantify the edge you have. For instance if the market says the odds are 10, and you think the correct price should be 11: lay it at 10 and as soon as the market goes to 11 (plus something to compensate for commisions) hedge your position. If it doesn't then you will have to wait and accept the huge profit or loss. This approach will lower the variance of your returns, because there are fewer outliers. In theory you should have higher return long term, because the money you just freed up can be used to lay at 10 again.

However, you might not like it because of psychological reasons. If you have a big loser from a bet you weren't able to hedge then the chance of you having a profitable day are diminished. Because there are fewer opportunities to win big again. In that case: accept the current variance and take really good care not to overbet.

[rinconpaul]

Appreciate the input thus far, but I think you're looking at the issue with a micro perspective. I'm talking global longterm. Let me see if I can impress on you my perspective:

If after 100,000 races of laying ($1000 liability) every horse in barrier 2, your balance has grown by $24,000. So there's an edge there. However if that took 7 months of races, 3 of them were losing months. Like you said, you can just wear it with the prospect that it'll come good?

Would a better hedging solution be to Lay barrier 2 and back the field in every race, but hedge your liability somehow such that you breakeven if the horse from barrier 2 wins, and profit if it doesn't?

[marksmeets302]

If you need 100.000 races for that my gut feeling is it might very well be a statistical fluke. Maybe it's better to check that first?

[gazuty]

If you need 100.000 races for that my gut feeling is it might very well be a statistical fluke. Maybe it's better to check that first?

Yes there would be a number of things to check. Is it indeed the signal or the noise?

I know that barrier draw is only being used as an example. But if it were true effectively what you have found is that the market incorrectly weighs the barrier in determining true odds (ie you have found an edge). That could well be the case. Barrier draw is likely to be relevant. Patterns could be found that were entirely irrelevant, eg. the birthday of the jockey. Humans are very good at pattern recognition but not quite so good at sorting signal from noise.

Before deciding anything is a signal I favour having a theory as to why it is relevant and how it can effect outcome. Have a theory as to how you have found a signal.

On hedging, you would need to factor in the cost of the hedge.

Taking your example, you have a $1,000 liability on the horse in barrier 2. To take an example let's hedge that at a $5 bet @ 200 in play. What are the outcomes.

1. Barrier 2 horse loses and the $5 is very very likely matched. You have reduced your winnings on the horse that won by $5.

2. Barrier 2 horse wins and the $5 is matched. You hedged and the result is you reduced your loss from $1000 to $5.

3. Barrier 2 horse wins and the $5 is not matched. You lose $1,000. This was your position in any case and reflects the status quo.

You would need to run the numbers and understand how much your hedge costs you. Case 1 and 3 seem most likely, but who knows until you look at the data. You need outcome 2 to occur in more than 1 in 200 races. Otherwise you are paying too much for your hedge.

So the cost of the hedge essentially pushes you to considering whether the hedge bet is also expected return positive. If it is not and if it is expected return negative then all that will happen is the $24,000 profit reduces. If backing the barrier 2 horse at $5 for 200 in play is long term more than -$0.24 expected return negative then your hedge turns the entire strategy negative.

Finally, if you graph the pure result is the saw tooth very slightly up? Or does it jag below the profit line into loss quite often. ie can you determine a trend ever so slightly upwards in the graph? If not that weigh on my thinking.

Overall I d think you'd be better off sticking to the pure signal (unless the hedge is positive or only very slightly negative). But, to take a Keynes quote out of context, in the long run we're all dead. (I'd say you were also thinking of his quote about markets staying irrational longer than one can stay solvent).

[rinconpaul]

I've been researching this idea using 3 years data and have come up with some promising results. Have any of you heard of 'Pairs' trading? It's a form of arbitrage used on stocks, futures and options markets. It's where you take opposing positions on stocks/commodities/indexes which in some way are related to a common performance. Like say CocaCola and Pepsi shares, they're both soft drink manufacturers and should be perform similarly? It's when one outperforms the other, you short one expecting it to return to the mean and buy the other.

Now if you group horses with similar traits, rankings etc...some uniform measure that's easily calculated, and over a long period for every race record the profit/loss if you were to lay and back them at Starting Price, you come up with 3 results:
1/ One side will profit by more, say backing then laying and vice versa.
2/ Both sides will lose backing or laying.
3/ Both sides will profit backing or laying.

In the case of #1 you would bias your stake toward the profitable side, using the other side as a hedge.

In the case of #2....forget those selections.

In the case of #3, it seems unusual, but it's to do with the variance/unpredictability of SP and therefore implied probability versus actual probability can do weird things. None of us know what the SP will be, but using the right classifications for your selections, you find that SP favours some sides of a trade more than others, neither side, or performs outside the deviation wildly either way. Because we never know the SP, you can't have balance stake/liability so I've based my research of 100 stake/1000 liability.

You might ask how often these outcomes occur?
#1 about 25%
#2 about 66%
#3 about 9%
As a long term exercise it could have merit, as you're taking a global perspective and trading this way in every race.

[rinconpaul]

For those interested in how this subject is developing: http://www.propun.com.au/forums/index.php?
Read my thread "Punt like a Hedge Fund"