Horse Racing - how important is specialising to be successful

There is value in specialising but there's a limit on how much you can specialise without creating huge holes in your data or without having very few opportunities to bet.

For example I specialise on National Hunt chases between Oct - End Apr. I don't do Hunter chases but do all classes. I end up having 500-600 bets over the period so say 80 a month on average. Now because I don't study hurdles I tend to avoid novice chases until they've run 1 or 2 times and because I don't follow summer jumps as soon as the ground starts firming up in late April and early May I have draw stumps on the season. But at least during the period I'm betting i'll have ratings on all the races I bet on.

If you specialise on a particular distance say sprinters on the flat, how are you going to assess a horse dropping back in trip from 7f or a mile? How will you know if the time of a particular 6f race on a given day is any good if you aren't able to compare it with the mile race on the same card etc. Specialising on tracks seems ludicrous to me unless you are limiting yourself to all-weather and even then you are going to be betting with very limited data and even more limited opportunity.

I know people who try and specialise on grade 1 races etc but then again they miss some of the best bets when an underrated handicapper comes up against overrated grade 1 horses.

So all in all I think its a good idea to specialise but I don't think its possible to cherry pick a very limited band of races.

One tip - don't try and specialise in 2 year old races unless you have an extremely good eye for the improver and knowledge of how differently trainers bring their horses along - its an almost impossible game IMO.

I spent hours looking over my results and quickly found my niche for each distance on the flat.

It helped immensely.

5 furlong sprints. On good to firm or firm, I was able to find solid EW plays.  Any other goings, not a chance.

6 to 8 - solid, steady returns.

10 to 12 furlong races - good and good to soft, solid returns.

Races under 60, look for a solid EW play.

etc.

That doesn't mean I don't look at other races.  It means this is my bread and butter and I put the time in to find the key bets for the day.

But don't stop there.  Look at the tracks where you do and don't have the ability to read and put a line through them.  Why put time and swim against the current.

Focus your time.  it is valuable.

I break down my stats to class of race, distance, track and many many other variables annually and have data back for 11 years now cumulatively. To be honest the results vary so much year to year - one year may be great for class1's, the next for class 2's, one year great for 2m races and bad for 3m2f+ and the next year that may reverse.

I think there is danger in interpreting too much into a season or two of data - for me that would be 1000 - 1200 bets. eg I found for many years class 4's were my best and 2m hcap chases were my best type of race whereas this year I lost in both categories. Using results data IMO is valuable to plan future specialising strategy IF and ONLY IF:

1 - You have enough data to back it up - think 3000 - 4000 bets should be fine.
2 - You have compiled a reason(s) for why your are more successful in a given area. eg: one or two big wins can skew data.

I have one active example of where back data has been useful to me in that January has always proved to be a tough month for me and is the only month I have lost at cumulatively over 11 years whereas the other months are showing 25%+ profit to stakes. Firstly I tried to come up with theories on why this is the case as I'm using the same ratings database etc and came up with a few theories that whilst I'm not 100% on make enough sense for me to feel there is higher volatility for me and my ratings. Only after this I have decided to scale back my betting in Jan.

If you take any data set and divide it up into sub categories you will find clustering, it can be very difficult to tell the difference between a meaningful grouping and normal but random variation. If you go through your results looking for areas where you won and lost you'll certainly find them, but take care - all gambling looks incredibly successful if you disregard the losers.

I have data running into several million transactions and I still have great difficulty confidently answering very simple questions like "do I generally do better at shorter odds or longer odds?"

In general "fishing" in your data and then coming up with a theory behind what you found should be avoided, a slightly healthier approach is first to establish a hypothesis and then test it with your existing results, but the only really meaningful test is to make a prediction before you have data, then test it. The trouble with this is that to do it properly you've got to be willing to provide yourself a control on your experiment. Say you expect to do well in shorter races - how do you know unless you also collect results for longer ones?

If I was giving advice on someone new to the horserace betting game . I would say learn to think like a trainer, follow certain trainers not neccassarily those backed by big money but new up coming ones.

One of my best bets this flat season was Great Page which won first time out at Windsor trained by Hannon and owned by Middleham Park I read their website and was convinced they thought a lot of it. It won at 3-1 not a big price but I thought it would be 6-4.

I am currently using an angle in on trainers who do well at certain courses where they are only running 1 horse on the day.

There are lots of different angles in but I would not specialise at one particular type of race.

Agree with alot of the above.

I use to look for horses in handicaps that where weighted out of the handicap but had previously ran at a higher RP rating.

I followed this theory for the best part of 3 years; it was a disaster.

Most of the horses I backed turned out to be on a massive decline, with very few exceptions.

Greyhound betting and the odd football lay kept me going, my horse racing history was shocking during this period.

Dav - I think your above theory comes from believing horse racing is all corrupt. On the whole I think you're better off believing what you see as the whole truth. Neither is completely true but I think believing what you see is real is the lesser of two evils than believing everything is corrupt and all the trainers are planning a plot.

Every day I see so many donkeys that go off at ridiculous prices because people somehow believe because they won off a 10lb higher mark 2 years back they are gonna come back to form despite the fact they've run like $hit for the last n occasions. Anyway I think the topic is digressing of specialising. OP doesn't seem interested so whilst I have a ton of views on specialising I think I'll knock this thread on the head.

I think a lot of bookies have got it wrong in that that have punter mentality in that they are looking for the winner and trying to lay the rest whereas they should be looking for the ones they think will shorten and lay the rest. Its a skill and not many have it. Having said all this its a devilishly hard game for them nowadays.

Racing -

May sound a silly question but what is the least number of lays you would be willing to place  per year ?

If you had a lay strategy that had around 200 selections a year and made a profit of around 7% on turnover  would you see it as worthwhile just to stick to it and nothing else ?

cheers

One of the best threads I've ever read.

Amongst others, I'm interested in aye robot's (good film) quote:
"In general "fishing" in your data and then coming up with a theory behind what you found should be avoided, a slightly healthier approach is first to establish a hypothesis and then test it with your existing results, but the only really meaningful test is to make a prediction before you have data, then test it."

I'm guilty of fishing in my data and then coming up with a theory. From a scientific, or betting, point of view I don't see how you really establish a hypothesis without first having data.

My use of the word "hypothesis" might not be technically accurate - but I'll give an example to show what I mean:

You might think that jockeys who have had a run of consecutive wins get over-backed and are therefore value to lay (I'm not saying that's the case, it's just an example). That's a plausible hypothesis with some reasoning to it that you can test going forwards because it's quantifiable and it allows a prediction - if it's true then you can expect a profit from laying all jockeys who've had (say) 4 consecutive wins. If you can predict something and then demonstrate it then you have gold standard evidence for your hypothesis (or whatever the correct word is). Maybe that happens, great - but chances are it won't turn out like that so let's run the example forwards.

To test the winning run hypothesis you place a load of bets but it doesn't work out - your bets roughly break even before comm as you would expect if you placed bets randomly. If you're being smart then that should be that - hypothesis busted. Problem is that's not easy to take so you're going to be tempted to dig in the results for some crumbs of hope. If you look for it you will find a sub-set of winners in any set of results, so maybe now you start thinking that your hypothesis is right but "only in sprints" or some-such. Forget it - if the rationale for your hypothesis was sound then it should apply to all distances, you don't strengthen your case by reducing your data set.

Now maybe you give up on the winning run jockeys thing but you've still got the data, so you start fishing in it for something else and you see that you did really well on those horses that were returning from a break. Well now you've got a new hypothesis and it's tempting to imagine that you've got some evidence for it - this is a big mistake, in fact you have no evidence at all for the new "returning from a break" hypothesis.

In every set of bets there will be a winning group who's members have something in common but if the thing they have in common isn't what you were expecting it to be before you started then that isn't evidence of a real phenomenon. Before you started it was a dead cert that there was going to be some sort of group of winners somewhere in your data because that's what you would expect even in a completely random data set.

It's fine to use your old data to come up with new ideas, but it will almost never provide significant evidence for them. For practical purposes there is a simple rule; Hypothesis first, data second - otherwise it doesn't count. I have to add the caveat that you could theoretically find an extremely strong signal in your data in which case it might be meaningful but it's very unlikely and it's far more likely that you're fooling yourself about the strength of the signal.


A slightly different topic now but whilst I'm here and feeling a bit ranty; statistical significance. If you're not sure how good your data is - it's no good. If you need to calculate Chi or do T tests then the signal you're looking at isn't strong enough to turn a profit against commission. Anything good enough to do that shows up on a graph that a half blind eight year-old could understand. Ok I'm overstating a bit and there are occasions when those tests can be useful, but in general if you're doing it right you shouldn't need any maths that you didn't learn at GCSE.

Zealot - I don't really get involved with laying horses much as I know my strength is finding horses that are priced wrong in the morning. Sure I can see horses that will double, treble in price from the morning BUT there's zero liquidity on here so trying to chase the horse out is futile. Much better for me to just get the win bets on early at the bookies.

I have 6 years of stats for type of bet 2009 - 2015:

WIN - +25.94% to stake,
EW  - +33.79% to stake,
LAY - -2.49% to stake

Actually my best %'s to stake are for Ante-Post, Multi-bets and FC's (B365 fixed odd ones) albeit to much lower volumes. So from my point of view Laying is an infinitely harder game so if you are clearing 7% net after commission etc you are doing very well but I think you are selling yourself short on your talents as I think you'd be able to net a much higher rate of return backing assuming you can organise the infrastructure to be able to get on with the bookies especially early.

OP doesn't seem interested so whilst I have a ton of views on specialising I think I'll knock this thread on the head.


racingguru, interested in any of your thoughts on specialising, as it's defo a problem for me...i do ok but there are many angles to attack horse racing and i'm sure i'm trying to do too much at once, which is not clever given the amount of racing these days.

aye robot - thanks for detailed and eloquent reply.
Still trying to work out if I agree with everything you say or not.

I'm talking about footy not horses but don't think that makes much difference.

Certainly agree with everything about "your bets roughly break even before comm as you would expect if you placed bets randomly". That was a situation I was in, for some bets, and have recently ditched the "system" I was using, I wasn't tempted to "strengthen your [my] case by reducing your [my] data set", so entirely agree with that as well. I now use data combined from top European leagues.

"In every set of bets there will be a winning group ...".  Agree with that.

And having read the paragraph starting "It's fine to use your old data to come up with new ideas ..." about 20 times (I'm a bit slow at times) I agree with what you're saying. I think I would have worded it "collect data, analyse data, form hypothesis, test hypothesis".  That's the bit I was getting stuck on.

And finally also agree on your "statistical analysis" mini-rant. I calculate percentage of wins in a data set and from this the odds required to break even, I compare this to the average odds in the data set (I have about 10 groups which I allocate games to and the odds in each group don't vary much)and the first(break even odds)must be significantly better than the second for it to be a bet.

Thanks again.

Thanks for the reply racing.

I think I  would do ok at backing but to be honest I just haven't got the time . I'm out most mornings at 6;30 and home around 6 ish and with young kids at home it would be nigh on impossible to find the 2 hours , at least , id need to study the following days cards. I'd love a couple of hours each morning just to study the cards as I do really enjoy it .

cheers

Specializing is only part of any success to gambling. One must also master their chosen speciality.

Specializing is necessary in any
sport, but understanding the betting
margins is the most important.

I wd say something slightly different in that rather than specialising at certain distances I wd try to know the characteristics of all the horses likely to run against each other in good races (both races in which I am genuinely interested and races with a certain early betting volume). This means knowing about something like 1,000-1,250 horses, mostly good class on the level, and just half that number of older horses over the jumps.

Knowing how the horses may run gives you an in not just on the going but only likely track and pace biases that aren't always reflected in the betting. There are some horses that are 10lbs better chasing a strong pace, held up in traffic or given a lead on a rail. There are horses that will act on some sprint tracks depending on where the stands are, notwithstanding the track's notionally being straight. There are horses that will extend up a hill e.g. at Newmarket or Sandown to overhaul a leader in a way that can be apparent even if they haven't previously run at the track.

As ken says, more important than finding potential winners are the technical skills of getting on in the morning and understanding which horses and yards have public support, back horses when they have a chance (and when--this also extends to connections), and which races are likely to see the jolly tightening. It is much easier to read the market for major Festivals simply by using form and an ordinary sense of favoured horses, yards and jocks.