Judging a jockeys ability has long been an inexact science, with some widely different views on how effective certain pilots are. Some jockeys like Jamie Spencer polarize opinion, while others like Ryan Moore and Frankie Dettori are generally accepted as two of the best around. What forms these opinions though can often be little more than us remembering a close call that went their way in a big race, or what seemed a good ride when we had them backed. Likewise we probably unfairly hold it against some jockeys when things don’t go as planned the day we had the money down on them. In summary it’s very likely our judgement of jockeys is biased by a small sample of races that effected us in some form or other, and once that opinion is formed, what they do in 99% of their other rides will likely pass us by.
The problem with jockey analysis is it’s very hard to do it objectively. For example I’ve heard some pundits claim you can judge a jockeys ability by measuring how well their mounts do versus how they were expected to do by the market. They would for example say Ryan Moore had 100 winners last season, when the Betfair SP market would only have predicted 90, and conclude that shows he’s better than another jockey who may have had 90 winners when 100 was expected by the market. Of course it shows nothing of the sort, all it measures is whether the market over or underrates the significance of a jockey booking, and/or the effect that jockey will have on the horses chances.
Likewise using tables of how many winners they’ve ridden both overall, or just in big races, or even strike rates, tell you more about how connections perceive the jockey, or how loyal they are to that jockey, than they do about the jockeys actual ability. Some use more sophisticated methods like comparing how a horse runs with various jockeys, so if say on average a group of horses ran four pounds better under Ryan Moore than your average jockey, then they would conclude he was worth four pounds more than your average rider. Again this is hugely flawed as I’d wager most of that improvement is down to intent from the trainer on the day, rather than the jockey, in other words today’s the day the horse has been lined up for. For example the lad that rides work for the trainer twice a week and gets a few rides to reward that, gets replaced by a top jockey, the ensuing improvement clearly can’t be fully attributed to the new jockey, as it’s likely it would have run much better for it’s previous jockey that day too.
How then can we use data to analyze jockeys? While jockeys often get the headlines for a strong ride that results in a narrow win, strength in the finish between competent riders, will rarely make more than a length or so difference in say a mile flat race, but getting the pace wrong can cause huge swings in finishing position. I’ve studied sectional times for over fifteen years, and the difference in a horses performance when they are paced optimally, versus when they go too quick, or too slow is stark. It seems prudent then to use sectional times to judge jockeys on the most important part of their job, judging the pace.
For this article I’m using the sectional timing archive from Timeform. They have one split per race, normally for the last two or three furlongs and also an overall finishing speed percentage, and upgrade figure. The archive starts at the end of March 2015, with the upgrade figures starting in November that year. It is possible to come up with more advanced methods of analysis especially with furlong by furlong sectionals, which are now available on the At The Races website, but for the purposes of this article the finishing split is more than enough to delve quite deeply into which jockeys have their horses better positioned when the race reaches the closing stretch. To learn more about the methodology behind sectionals upgrades I suggest you follow this link to a page with a free download of Simon Rowland’s Introduction to Sectional Timing.
In this article I will look at some of the bigger name jockeys, and go through various methods of evaluating their pace judgement, and in a follow up article I will do charts and analysis using the same methods to pinpoint other notably positive or negative returns from other jockeys. I use Proform Pace figures to ascertain how the horses were ridden. They are split into Led, for led or disputed, Prominent if the horse was positioned in the first half of the field early on, or held up for horses positioned from the middle to the rear.
I think everyone will agree that the place to be in a slowly run race is at or near the front, while in a race when the leaders are going too quick, it would be a much better use of your horses energy if you sat off that pace a bit. Granted some horses have some rigid preferences, like needing to lead to win, but most are more adaptable than that. I heard Ruby Walsh say more than once that he let the pace decide his position within a race, and you would assume most top jockeys would do the same. In the above chart I’ve split races into Fast or Slowly run, and then checked to see what percentage of the time each jockey led for each. I first came up with average race finishing speeds for each distance/track, and then used anything over 2.5% above or below this average as fast or slowly run. If the average finishing speed was 102.5%, then less than 100% would be a fast run race, as the runners are finishing slower relative to their overall speed, and over 105% would be slowly run as they are finishing fast relative to overall. This resulted in a sample of about 17% of all races for fast, and similar for slowly run.
The first thing to note is Frankie Dettori led the highest percentage of times in fast run races, and more significantly led more often when the pace was faster, which is hardly ideal at all. He was closely followed by Silvestre De Sousa, while Harry Bentley comes out well here as he very rarely leads when the pace is quick. Jamie Spencer is another who rarely leads when the pace is quick, but as can be seen by the orange line, he rarely leads at all, as he has a similarly low percentage when the pace is slow. De Sousa clearly rides his horses more prominently than most, as he also leads quite a bit in slowly run races, but worryingly that percentage is still lower than for fast run ones. Jockeys will have their riding styles, and as such some will lead more often than others, but a good jockey should be more likely to lead when the pace is slow, than when it’s quick. As can be seen from the above charts, Donnacha O’Brien comes out of this particular test with flying colours, as his orange line far exceeds the blue one, leading 25% of the time when the pace is slow, but just under 9% when they are going a good clip upfront. Considering he is much younger and has less experience than most in this sample, that is very commendable indeed, He clearly does his homework before a race and is ready to dictate a steady gallop when the opportunity arises, and it could pay to note races he rides in with little obvious pace, as he seems very adept at taking advantage. Ryan Moore and Jim Crowley also seem far more likely to lead when the pace is steady, than quick.
Dettori is actually third highest at leading when the pace is slow, but he still leads less often than when they are going quick, and I can only surmise that when he decides to lead he does it regardless of how quick he needs to go to accomplish it. Daniel Tudhope, and William Buick are others who come out poorly as their blue line exceeds their orange one, when it should be the other way around.
The above chart shows how often each jockey held a horse up in a slowly run or fast run race. Unlike the charts for leading, it would now be ideal for the blue line to be greater than the orange one, as this indicates they held up a higher percentage of horses in races run at a decent pace, than when the pace was steady. Jamie Spencer has easily the highest percentage of hold up rides, and he was only marginally more likely to hold one up in a slowly run race than a fast run one. Donnacha O’Brien again comes out very well on this metric, with his blue line greatly exceeding his orange one. In races run at a good gallop he held his mount up 57% of the time, but when they pace was steady he only held up 37% of his rides, again this shows he either had his homework done before the race on the likely pace, or adapted quickly in the race. Ryan Moore also comes out well on this measurement, as does Jim Crowley and Andrea Atzeni. Seamie Heffernan however, holds up around 55% of his rides in the sample races, and it doesn’t matter what pace the leaders are going, as he is as likely to be near the rear if they are going too steady, or too quick. Silvestre De Sousa and William Buick don’t come out much better, as they too don’t seem to let the pace of the race increase or decrease there likelihood to hold one up.
To get an idea of a jockeys overall style I came up with a pace number using 1 for led, 2 for prominent, and 3 for held up. I then got an average for all of a jockeys rides. The figures will naturally be over 2 as half the field are counted as held up, but only about 10% lead. Based on the earlier charts it’s hardly a surprise that Jamie Spencer has easily the highest figure, as he holds up much more horses than he leads on. Colin Keane, Harry Bentley and Seamie Heffernan also tend to ride their horses more towards the rear. Silvestre De Sousa, Frankie Dettori and William Buick are at the other end as they tend to ride their horses more prominently.
Now that we have pace figures for each jockey, with a higher one meaning more likely to be held up than ridden prominently, we can use these to test each jockeys positioning in fast or slowly run races. They have the benefit of joining the previous led/held up tests into one metric. Jamie Spencer as his blue line indicates does tend to hold his horses up further back in slowly run races, than quick ones, but it is only marginal, and he is also by far the most likely to get caught out the back in a slowly run race. From previous charts it will be no surprise to see Donnacha O’Brien has the biggest gap in his overall positioning of horses depending on the pace of the race, with Ryan Moore doing next best in this regard. Honorable mentions also for Jim Crowley and Andrea Atzeni. Daniel Tudhope, William Buick and Seamie Heffernan don’t do so well though, as their positioning within a race seems to be the same regardless of pace, and they’re actually fractionally further back in slowly run races.
Perhaps a more sophisticated method to gauge a jockeys ability to judge pace effectively, and to use their horses energy optimally, is to use the Timeform Upgrade figures which are explained along with the formula used to calculate them, in the article I linked to earlier in this post. Neither I, nor Timeform would claim these figures to be facts, the upgrades are just an estimate of how much quicker a horse could have run if it was paced optimally. Optimal finishing speeds for each track and distance are arrived at by collecting a sample of horses that ran a fast time relative to their ability, and as such are likely to have raced close to optimal pace. Horses are not exactly alike and have different preferences, but I can testify that from hundreds of hours studying both overall times, and the sectionals that led to them, that the differences are quite small, and as such if say optimal finishing speed for an average horse was 102%, then even a horse who prefers to expend less energy early, and more late, will still run it’s fastest time when it’s quite close to that 102%, it might be 103% but it won’t be much more.
For proper analysis of a race upgrade figures should be used in conjunction with your own video and form analysis to spot cases when they are likely to mislead, For instance say a sprinter steps up to a mile, and the pace is a crawl for the first three furlongs, but he settles well in the rear. He then makes plenty of ground in the final two furlongs to dead heat with the horse who tried to make all, and who battled back when challenged fifty yards out. Upgrade figures would give a big mark up to the sprinter, and suggest he would have won easily off an even gallop, when the truth is he very likely wouldn’t have stayed, and probably wouldn’t have figured at all, and it was only the fact he settled so well, that meant he could use his speed at the end.
That extreme example aside the upgrade figures are in my opinion a much better estimate of what would have happened if each horse ran their race optimally than the bare result alone. To use them to evaluate jockeys it’s no good averaging the amount of pounds each jockeys horse was upgraded, as a jockey ridding a horse who could have run four lengths quicker, that finished last, likely didn’t do much wrong, as the horse just faded having tried to win the race, but didn’t have the ability to do so. On the contrary though if a horse finishes fourth beaten three lengths then the very same upgrade means the jockey likely got the best horse beat.
I first calculated the amount of actual wins each jockey had, and also their estimated wins having accounted for the upgrade each horse would get, and then in the chart above I list the difference between each jockeys actual wins and the amount of wins estimated if every horse was ridden optimally. So say Horse A, with an upgrade of two lengths, won by one length from Horse B who had an upgrade of four lengths, and was two lengths ahead of Horse C who had an upgrade of seven lengths, the new expected finishing positions would be C from B from A. The advantage these charts have over the earlier ones is we are now using all races, instead of just fast or slowly run races, which only amounted to a total of 34% of all races, although that said fast or slowly run races lead to bigger upgrades than averagely run ones, so they would still have a big influence on the overall sample.
The above chart is actual wins versus estimated wins after each horse was upgraded to estimate how they would have done if paced efficiently. Oisin Murphy comes out of this looking good, having ridden 23 more winners than would be expected if all the horses in the race were paced optimally. Ryan Moore rode 19 more winners than expected, while despite coming out poorly in the metrics we used in fast and slowly run races, Silvestre De Sousa actually shows good pace judgement when all races are included, and more subtle effects can be measured within a race as well. Dettori, Crowley and Kevin Manning also do well, and while Donnacha O’Brien has only ridden 10 more winners than expected, he has done so from a smaller sample, so it is still commendable.
I’ve now used the same method with the upgrade figures except this time its for places. For the purposes of this piece a place is finishing in the first three. The above chart shows actual places versus estimated places. Harry Bentley comes out on top here, with 23 more places than the model predicted if all horses were ridden optimally. Harry actually had a very small negative figure for wins and the most likely reason for doing better with places is his ridding style, which as his pace number shows tends to edge towards being held up. This style will often see horses optimize their place chance by keeping going at an even pace, when horses who were in the heat of the battle falter late on. Jim Crowley has done well on every metric presented and he again comes out well here with 18 more places than estimated. William Buick does better on the place metric than the win one, and this is less easily explained as he tends to ride horses more forward. Variance will likely have played a part, or maybe he’s just one of the better jockeys for keeping a horse going for a place, when some others tend to ease off once beat. Silvestre De Sousa does much worse for places than for wins, and it’s likely his aggressive ridding style is the reason, which is fine for horses that are good enough to win, but means plenty of the others hit a wall that can cost them a place.
The differences between actual and estimated winners and places over a period of almost four years, may seem quite small if pace is such a big part of what decides a race, and some of the findings wouldn’t be that significant. Because one sectional provides less of the overall picture, it means adjustments have to be on the cautious side though, and if we had furlong by furlong sectionals available for the analysis, the ensuing model would be able to make much more adjustments in predicted finishing position. Having only one sectional means it’s harder to pick up more subtle variations from optimal pace. For example if the sectional split is for the last three furlongs, and the jockey kicks for home just over three furlongs out, does two quick furlongs but a very slow last one, the sectional will show the final three furlongs were even pace versus the full race, but in truth he went too quick from the three to the one pole, which meant the horse didn’t get home.
On the metrics used in this article Ryan Moore, Donnacha O’Brien, Jim Crowley and Oisin Murphy do very well, While on the opposite side it would be impossible not to conclude that the pace of the race seems to have little effect on Jamie Spencer’s ridding style. He’s still a great lad to have on your side in a big field handicap at Ascot, but perhaps he would do better if he wasn’t so rigid in his placement of a race within a race. He is typecast as a hold up jockey now though, and perhaps the horses he gets to ride can’t be ridden any other way, but it’s hard not to conclude he gives more than his share too much to do. Seamie Heffernan and Daniel Tudhope don’t come out of this analysis very well either, with Seamie in particular doing poorly on a number of metrics.
This article only concentrates on a selection of the big name jockeys ridding today, but there is many more interesting nuggets to share from the data, which I will do in a follow up article. It may interest you to know the jockey who comes out best of all for his pace judgement is not in the fifteen jockeys covered so far, so please check back in about ten days for more.
If you want my best bets everyday, check out my Premium Advisory Service. The profit since launch until the end of August 2019, for the Premium Bets and the Festival Packages combined, which Premium Members also receive, is + 1,946 points, which is +€48,650 at the recommended €25 a point stakes, for an exceptional ROI of +17.8%. The figures quoted don’t include any BOG either. The recommended starting betting bank for €25 a point stakes is €3,500, so that’s a remarkable 1,390% bank growth in just five and a half years.