Cut Your Losses?? (Mean Reversion)
Reversion to the mean is generally applied to much longer time frames than the short life of a specific horse race market. How do you determine the mean of a distinct horse race market where the bulk of the activity is close to the off. A steamer is not likely to revert to the mean nor is a drifter. As I said before, suicidal policy and with no substance or foundation for horse race markets.Ferru123 wrote:What's your reasoning behind that belief?
Jeff
I made an exhaustive study on reversion to the mean for greyhound trainer statistics and trap statistics covering every BAGS race held in the UK for the past 18 months. The study did not yield any evidence that betting candidates due for mean reversion would be profitable.
Thanks for your reply Dogform.dogform wrote:A steamer is not likely to revert to the mean nor is a drifter.
I don't know whether it's 'likely' to revert to the mean, but trends often break down and reverse.
I'm not sure I follow. Are you talking about situations where you bet on a steamer or a drifter reversing? If so, what were your parameters?dogform wrote:I made an exhaustive study on reversion to the mean for greyhound trainer statistics and trap statistics covering every BAGS race held in the UK for the past 18 months. The study did not yield any evidence that betting candidates due for mean reversion would be profitable.
Also, I think what applies to greyhounds may not apply to horses, given that the liquidity in greyhound races is much lower and you therefore possibly have different players in the greyhound market.
Jeff
Ferru123 wrote:....I'm not sure I follow. Are you talking about situations where you bet on a steamer or a drifter reversing? If so, what were your parameters?
Jeff,
The application of reversion to the mean that I analysed for greyhound racing was to calculate daily the trainers winning percentage for their last 100 runners and create a daily evolving trend line. I then compared this trend line to their overall winning percentage for the previous 6 months which was used as a baseline.
By comparing the deviation of the trend line to the baseline I then checked what would have been the P/L by backing/laying the trainers runners whose percentages were quite a way from the mean and would be due some type of reversion.
I did the same thing for the trap stats and in both cases there was no significant P/L. The reversion to the mean does take place but not in a predictable pattern that can form the basis of a system.
Regards
dog
Yes, it seems we were at cross purposes to some degree. However, I still maintain that reversion to the mean in a specific race market cannot be guaranteed. The market will have its mean which can be calculated on a split second basis using the correct data capture and analysis model, however, to then use the current deviation from that mean as the basis for trading is fraught with risk.Ferru123 wrote:Thanks Dog. I think we were talking at cross purposes. I'd thought you were referring to mean reversion in the sense of a going against the trend when trading in the hope that the market might revert to mean.
Jeff
It is in this light that I meant that a steamer will not revert to the mean, as it actively steams it will continue to diverge from the mean.
Regards
dog
Jeff don't know if this will help as have not tried trading at random exactly, but a few years ago i analysed my trading data and found out that something like 98% of my trades would have made a profit of some sort regardless of whether the initial bets placed were a back or a lay.Ferru123 wrote:I've found this thread very interesting.
Has anyone else experimented with random trading, where you enter at random with a stop of x ticks and an offset of y ticks?
Jeff
so to me that tells us that trading with a random entry point is very possible if we are able to spot a good exit point, if that makes sense
I agree with Euler that reversion to the mean needs to be measured at different timescales. I'm not an expert at trading horses, however many times you will notice that the ladder will display a skewed distribution, therefore no reversion to the mean. However at a smaller time scale eg 10 min pre off you may find it to be different.
See Re-Examining the Hidden Costs of the Stop-Loss for technical review!
The markets have changed a lot since the original post four years ago. However, I still dont use stop losses as such (just mental ones)
The ebb and flow of the markets are significantly different too. You can see the traders money hit the ladder now and more often than not you can see corresponding closing trades too, i.e. their money comes it, pushes the price in one direction and then they trade it out (and often push it back to where it stated). I know its always been like that, but it's more pronounced now
Regards
Peter
PS When I look back now to the original post so long ago,I learned a lot from that experiment (and all the subsequent ones). I still run tests each day to 'keep my hand in' as the markets change so fast and when real fundamental changes do occur (i.e. Cross matching) you need to ready to capitalise on them. The six months following Post cross matching probably represents half of my lifetime earnings on betfair.
The ebb and flow of the markets are significantly different too. You can see the traders money hit the ladder now and more often than not you can see corresponding closing trades too, i.e. their money comes it, pushes the price in one direction and then they trade it out (and often push it back to where it stated). I know its always been like that, but it's more pronounced now
Regards
Peter
PS When I look back now to the original post so long ago,I learned a lot from that experiment (and all the subsequent ones). I still run tests each day to 'keep my hand in' as the markets change so fast and when real fundamental changes do occur (i.e. Cross matching) you need to ready to capitalise on them. The six months following Post cross matching probably represents half of my lifetime earnings on betfair.
Blimey, 4 years ago
Things have changed significantly since then, but I stand by original disapproval of stop losses.
We have seen BA automation go to a new level since then, so a stop loss is obviously a key part of an automated trade. However, if you are watching a market, a good trader should be able to rescue a situation manually far better than an automated stop loss
Things have changed significantly since then, but I stand by original disapproval of stop losses.
We have seen BA automation go to a new level since then, so a stop loss is obviously a key part of an automated trade. However, if you are watching a market, a good trader should be able to rescue a situation manually far better than an automated stop loss
peterle - sorry for digging this up sooner than you had forecast. however, it is interesting to note that market efficiency even back then (2011-2014) had all but negated the stoploss. as we stand today, i'd say that the efficiencies are such that there is no way on earth that one could calculate the *best* stoploss methodology generically, without entering into a labyrinth of mishap and loss. have to agree with leTiss in an earlier post regards using gut feel and experience to guide exit.. sure, it means baby sitting the markets but if there was a one button approach, we'd all be there (peterle and pw excepted of course!! ).
so in true ronnie corbett style, i unwittingly arrived here (and digressed in my story above ) as i was actually implementing vwap into a sneaky little test piece that i'm working on - funny how the interwebs always brings you back home in a ridiculous and circuitous fashion
so in true ronnie corbett style, i unwittingly arrived here (and digressed in my story above ) as i was actually implementing vwap into a sneaky little test piece that i'm working on - funny how the interwebs always brings you back home in a ridiculous and circuitous fashion