|Welcome Guest ( Log In | Register )|
Articles >> horse-racing >>
The Bounce Theory
The Bounce Theory
This week I am going to try and investigate the “Bounce Theory”. It was a theory initially brought to our attention by an American horse player Len Ragozin. His idea centred around the fact that if a horse ran a career best time or produced a much improved career best performance, there was a strong chance that the horse would ‘bounce’ next time. ‘Bounce’ meaning the horse suffered a negative reaction to that supreme effort and hence tends to flop or at least run well below form on this very next run.
The whole “Bounce Theory” has been expanded by other race commentators over the years with an example being if a horse runs three improved ‘figures’ be it speed figures or collateral form figures, then there is a strong chance the horse will ‘bounce’ on their next (4th) run. In Britain the “Bounce Theory” has tended to centre around horses running well after a long break and then ‘bouncing’ if returning to the track quickly. It is this idea that I am going to focus on in this article.
One of the problems of testing this particular ‘bounce’ idea is deciding the parameters for the research. Questions such as, ‘how long a break should we include as a starting point?’, ‘what length of time do we use for a quick return on their second run?’, ‘can we group National Hunt horses with flat horses, or should we split them?’, ‘how do you determine running well after a long break?’, etc, etc. These are problems that people have struggled with over the years, and perhaps why little ‘concrete data’ is available in connection with the “Bounce Theory”. From my point of view, I am simply going to choose the time frames and performances as I see fit – whether they are ultimately ‘accurate’ or not I don’t think anyone can confidently say.
First things first I have decided to split the research into two – National Hunt racing will be my initial focus, before moving onto flat racing. I am hoping both codes will show similar results.
I have decided that a break of over a year (366 days or more) is my criteria for a long break. Looking at my data it can be seen that not surprisingly, National Hunt horses that have been off the track for over a year perform poorly. Despite improved training methods in recent years, a very long break is invariably a negative. Only 365 of the 6178 runners managed to win after a break of over a year – this equates to 5.9% of runners. Indeed, only 10.2% of runners managed to win OR finish within 4 lengths of the winner. A massive 36% of horses failed to finish (28% were pulled up).
This data clearly demonstrates that any horse that has won, or indeed finished close up (within 4 lengths) after such a long break had run well. These seem to be the horses to concentrate on when testing the “Bounce Theory”. They give us a decent enough sample size (618 runners) to test.
The next question is how soon should their next run be to test the “Bounce Theory”? 3 weeks? A month? 6 weeks? Again a difficult one. I decided to look at all 618 runners and their performance next time to give me a benchmark from which to work from. Then I would split the ‘days since last run data’ up to see if horses returning to track more quickly under-perform or not.
Here are the figures for the follow-up run:
First point to note is that 31 horses have yet to run again – maybe their long lay-off was connected with an injury or problem that has resurfaced. Focusing on the raw stats a strike rate of 18.2% looks about the figure you would expect for horses that either won LTO or finished within 4 lengths of the winner regardless of time off the track. The question now to be asked is ‘do horses returning to the track again ‘quickly’ ‘bounce’ or not?’ Deciding on the ‘days since last run’ grouping is not straight-forward. I decided on five groups – 14 days or less; 15 to 28 days; 29 to 42 days; 43 to 80 days and 81 or more days. Here are the findings:
The results are not what the “Bounce theorists” are advocating. Horses returning to the track quickly have out-performed the other groups. Indeed a 25% strike rate for the 14 days or less group coupled with a tiny profit seems to totally buck the theory. In addition if you look at the strike rates of the groups, the shorter the break the better. It should be pointed out at this juncture that the 43 to 80 days group also shows a profit, but in truth the strike rate is moderate and the profits are down to a couple of decent priced winners, which in a small sample of 60 runners can skew the figures.
As with the National Hunt research I have decided that a break of over a year (366 days or more) is my criteria for a long break. However, I am only going to include winners and those that have run to within 2½ lengths as opposed to winners and those that ran within 4 lengths. My reasoning being, that flat races are shorter, and hence a closer margin was required when determining a good run. I also used 2 extra years of results to give myself more data. As with the National Hunt results, flat horses returning to the track after a year or more performed poorly scoring just 5.3% of the time with losses of 42 pence in every £.
As with the National Hunt data, any horse that has won, or indeed finished close up (within 2½ lengths) after such a long break had run well. So once again these were the runners to concentrate on.
Here are the figures for the follow-up run:
First point to notice is how much poorer overall the results are for the next run on the flat as compared to the National Hunt results. Let me now split these results up by days since last ran – I have used the changed the groupings slightly combining any horse off the track for more than 4 weeks together. The reason for this is most of the flat horses returned to the track within 4 weeks (70% in fact).
These figures are the reverse of the National Hunt ones. It does seem flat horses after producing a good run after a long break, perform very poorly if returned to the track too quickly – in this case within 14 days. The betting market seems to overrate these quick returners as the losses are a hefty 60 pence for every £ wagered.
So why are the National Hunt and flat ‘bounce’ results so different? The logical explanation would be that flat bred horses are more fragile than National Hunt bred horses and hence being asked to `repeat` such an effort after such a short space of time is ultimately beyond most of them. This theory cannot be proved, but it seems the most likely reason. Indeed, I decided to look in more detail at the 16 horses that managed to buck the flat ‘bounce’ trend – eg. the horses that won when returned to the track again within 14 days. My hypothesis was that these horses would be worth following as they had shown a level of performance (within their class) that very few flat horses could match. The results from this albeit small sample were encouraging – if you had backed these 16 runners on their next 2 starts you would have won 10 bets from 29 for a profit of £7.48 (ROI +25.8%). So assuming you bet £25 per horse this profit would equate to £187. Not a bad return.
Although this article has only really scratched the surface there are certainly some interesting findings. One could argue that the idea should be tested using a different time span and/or different criteria in terms of what I have judged to be a ‘good run’. Just for my own peace of mind therefore, I decided to see if a longer break initially on the flat made a difference. I kept the ‘good run’ criteria to either winners or those that finished within 2½ lengths, but extended the initial long break to 500 days or more. I then focused on those horses that returned to the track quickly on their next start (within 14 days) – in other words the ones that were most likely to ‘bounce’. The results, despite being a relatively small sample backed up the theory – of the 52 qualifiers only 2 managed to win within this 14 day period. Indeed, they were fully expected to win being priced Even money and 5/2 respectively. Anyone backing all 52 runners would have lost a massive 89 pence for every £ wagered.
Copyright echo date("Y");?> PunterProfits.com