Lessons learned from a (debut) punting season in HK

I punted seriously for the first time ever on Hong Kong racing during the 2019/20 season, having produced my own speed ratings and sectional-adjusted speed ratings (SASRs) for all races run in HK in that timeframe.

As expected, it wasn’t easy going. HK racing is a different beast to UK and Irish racing. The same long-held beliefs central to successful punting closer to home needed to be challenged and, in many cases, discarded.

With all that in mind, I was delighted to have finished my debut season in profit (albeit having made modest gains, and with UK bookmaker incentives offered during the COVID-19-imposed blackout on home shores providing me with unusually generous opportunities).

With a promising debut season behind me I am looking forward to playing more seriously next year, with vastly improved speed rating and sectional analysis tools behind me.

Some key lessons learned from the 2019/20 season have driven these improvements, and I am going to go into more detail about these below:


1: Accurate standard times are crucial, and these change from season to season

An obvious statement of course, but when you are diving in for the first time it is hard to know how accurate your standard times are, and only after a lot of races do some small but significant issues pop out.

For clarity, standard times are important in compiling speed ratings because they allow you to predict how fast each race should be run on a given card. This is imperative in calculating what effect the going has had on times, and, by extension, in putting a standardised value on each performance. If some of your standard times are inaccurate, this otherwise simple but crucial step in compiling the ratings becomes more of a guessing game.

I initially compiled standard times by taking median winning times from all races run at a given course, distance and class (solely on going described as good) over the past eight seasons. However, as the season progressed, some times seemed too fast and some too slow.

For example, horses running over 1650 m at Happy Valley were underrated as the standard times for such distances were seemingly faster than what could reasonably be achieved last season. And those running over 1000 m at Sha Tin were overrated as the opposite was true.

Being aware of this issue meant I was able to factor it in to any decision-making process around my betting, but of course the big problem came when trying to compare speed ratings for horses running over such distances with those running over others. Was a 90 achieved in Class 3 over 1650 m at Happy Valley better, worse, or similar to a 92 achieved over 1200 m?

Of course, the whole point of speed ratings is that they standardise performances no matter what distance, course, class and going horses have been running over. A 90 should be a 90 should be a 90.

For the upcoming season (2020/21) I have decided to revamp my standard times completely by only calculating median winning times for each course, distance and class (on good going) for races run last season (2019/20), which, importantly, saw horses run significantly faster times than had been the norm at most course and distance options in previous years.

There are of course some gaps (or issues where data are scant for certain race conditions) but these can be (fairly) reliably estimated via models based on the differences seen between times at similar distances and classes where there are more data.


2: Weight changes do not have universal impacts on speed

It was clear from my ratings last year that major changes in weight (following a victory) do not have a consistent impact on subsequent ratings. Logic would dictate that carrying more weight should slow a horse down, but there are numerous cases where this might not be so (such as when a horse is improving rapidly or gets an improved trip in running).

Weight is also almost certain to impact different horses differently depending on their mass. Helpfully, the HKJC publish horse weights (lbs) before they race each time, so not only can you track changes from one race to another for each horse, but you can also observe the huge variety of masses across the whole population. Some horses are as small as 950 lb or so, while some are over 1300 lb. Clearly a 10 lb weight hike ought to impact the former more than the latter.

Even allowing for the differences and focusing on analysing horses falling within a fairly narrow mass band who suddenly experience major weight rises paints a confused picture from a data perspective. Generally, a slightly lower speed rating is predicted after a major weight rise, but there are lots of exceptions and the median reduction in speed rating is so small I am inclined not to factor it in to any predictive ratings I produce ahead of a day’s racing.

Note there is one major exception to the above rule: when a horse drops in grade (e.g. from Class 3 to Class 4) it will usually be dropping due to poor recent performances and will likely be shouldering a lot more weight in the lower Class (having been carrying low weights correlating to being rated towards the bottom of those running in the higher class previously, the opposite will now be true). This double blow seems to hit these horses hard and they routinely run considerably slower speed ratings when first in the lower class until finding a point at which they start to become more competitive. Of course. there are some who win at the lower level first up, but these are rare.


3: Pace maps / race shapes are crucial in impacting speed ratings

This is something that many will have anticipated but it is worth explicitly mentioning as it holds the key to assessing each and every race in HK.

Similar to racing in the US pace is ultra-important in HK racing, and – in tandem with the draw (discussed next) – the most important element in determining whether a horse will run to its full potential.

A universally applicable mantra to all speed ratings is that a fast horse can run a slow rating (if the pace is slow, for example) but a slow horse cannot run a fast rating.

In this context, it is important to understand that plenty of horses record speed ratings considerably lower than they are capable of. Sectionally-adjusted speed ratings should be capable of upgrading horses effectively based on the pace each race was run at, but this is in practice more complex than might seem obvious (discussed in point 5 below).

At a minimum, I have learned the importance of noting which horses were disadvantaged by the pace / race shape so that I am aware that they are capable of running higher speed ratings than those they have recorded on such occasions.

In practice, predicting how a race will be run is not an exact science, but more often than not it is possible to produce a workable pace map (factoring in the draw) based on the positions horses have typically taken up in the early stages of their recent races.

In other words, it is possible to predict with some accuracy how most races are likely to be run and (sometimes) pinpoint horses likely to suffer from the shape of the race (e.g. those likely to be held-up with inside draws; front-runners drawn wide in races with lots of speedsters; and mid-div horses likely to be forced to race three or four wide as they are drawn outside similar horses).


4: The draw is a huge factor in most races – but it is context-specific

This builds on what has been discussed in point three above.

In simple, generalised terms, it is a negative to be drawn wide in races up to 1650 m at Happy Valley (especially over 1000 or 1200 m), and in races up to 1600 m at Sha Tin (on the turf and all weather tracks, with the exception of the straight 1000 m track at Sha Tin (turf).

However, this is context-specific.

For example, it is generally a big negative to be drawn low in such races if a horse is a hold-up performer as it will typically find itself behind a wall of horses rushed forwards early on and then have to hope and pray for a gap to appear (they usually don’t).

Racing more than one horse wide is a cardinal sin in HK (especially at the tighter Happy Valley track) and so, another category of horses that often struggle are those that prefer to run mid-div and that have to break from middle stalls. Generally, they won’t have the pace or inclination to bag prominent spots and will either be forced to sit further back than desired or race three or four wide and be more prominent.

The last group of horses that commonly struggle to make the impact their ability deserves are the hold-up horses that find themselves in races lacking much pace. As you’d expect, these races often go the way of the few front-runners/pace horses in them who are able to dictate matters at an easy pace before quickening at the point they choose.

In all cases, a race-specific pace / race shape map will be helpful in flagging any obvious concerns for certain horses. The easiest way of putting these together is to take the median position each runner took up in the early stages of its recent races and lay these data over the draw to pinpoint which horses are likely to lead, sit mid-div, or be held-up and then flag those likely to have competition for the spots they want to take up.


5: Sectional upgrades require a race-specific (not a C&D par) focus

The current thinking (for UK and Irish racing) is generally (not everyone agrees) to calculate par finishing speeds for race winners over each course and distance by focusing on races that were fast run for the class.

The logic is that if the mean or median finishing speed for fast winners of a race over a given course and distance is, say, 105%, sectional-adjusted speed rating upgrades can be calculated based on a formula that awards bigger upgrades to horses that finish further away from that 105% figure than those that finish closer to it.

However, analyses in HK confirm that horses can run fast times for the grade in multiple ways, which strongly suggests there isn’t an obvious, magic-bullet finishing speed % that shows how to run most efficiently over a given course and distance.

Even more important in my mind is the idea that horses who run very different styles of race (as highlighted by finishing speed % and by distance matrix analysis) to those who finish in the first three in any given race are the ones that should be getting the biggest upgrades.

For example, if we use our hypothetical par finishing speed of 105% for a given course and distance, suppose the first three in a race over that course and distance ran finishing speeds of 102%, 103.2% and 102.4%, whereas the sixth home ran 106%.

That sixth horse evidently came home very quickly relatively to the rest of his sectional times and was thus likely left too much to do off a slow pace. Yet his upgrade would be less than the first three, whose finishing speeds deviate by more from the par of 105%, unless we take a race-specific approach.

If that assessment looks accurate from watching the race, a distance matrix will be helpful in confirming as much. These matrices compare how each horse ran each sectional of the race relative to its competitors. You can see an example below to help visualise this (in this example, the fourth home, V Chevaliers, came from a long way back and produced a finishing speed % much higher than those in the first three, who ran more even fractions closer to the pace).

My working hypothesis (to be tested more completely in 2020/21) would be to take a race-specific par finishing speed (the median finishing speed % of the first three) and then award any upgrades based on how far each horse deviated from this.

In such a scenario, the fourth home in the above example (V Chevaliers) would receive a bigger upgrade than all horses on the left half of the matrix (including the first three home), as he ran a very different style of race to those that filled the first three spots and was thus likely most disadvantaged by the way the race panned out.


Changes ahead for 2020/21…

Firstly, I am delighted to say I will once again be compiling my own speed ratings and associated data for the upcoming 2020/21 HK season and that I will be making these freely available once more.

If you haven’t already signed up to receive regular updates via email, you can do that by signing up to my mailing list from here

With many of the lessons learned from last season in mind, I am going to re-work a number of processes in the 2020/21 season. As a result, the speed ratings and database of other information I produce will see some changes and useful additions.

The most important are:

1: New scale for all ratings

Last season, I “centred” the scale for my ratings so that any horse running the same time as my (highest) Class 0 standard time – once the effect of the going had been taken into account – would receive a rating of 100. Horses running faster than this would receive ratings above 100 and those running slower would receive ratings below 100 in proportion to how much faster or slower they were.

There was nothing wrong with this arbitrary decision (at least in theory), but I now think it makes more sense to “centre” the scale based on how much faster or slower horses run in comparison to the Class 3 standard time for each course and distance.

The reason for this is simply that there is hugely more data available at this Class than at Class 0, and, as a result, I can be more confident in the accuracy of any median times calculated for this Class and, by extension, the accuracy of any ratings for horses running times that deviate from this mark.

In an attempt to keep the ratings awarded very similar to those last season, I will centre the scale so that any horse running the same time as the Class 3 standard time – once the effect of the going has been taken into account – will receive a rating equivalent to the median rating awarded to Class 3 winners last season (93.3).

In practice, this should mean that most winners in most Classes will receive similar ratings to most winners in most Classes last season (despite the change in methodology).


2: New methodology for sectional-adjusted speed ratings

As described in the fifth lesson learned above, I will be producing upgrades based on how far each horse’s finishing speed % deviates from the median of the finishing speed % of the first three home (and with the confirming help of distance matrix analyses). I believe this technique will more accurately reflect how disadvantaged horses are on a case-by-case, race-specific basis.

One caveat to this will be that I don’t envisage upgrading horses that have had the same chance to run big ratings as the first three have done — even if their finishing speed % numbers suggest otherwise.

A good hypothetical means of picturing what I mean here is to imagine a race in which those racing on the front end dictate and fill the first three spots. Say the median finishing speed for these three is 100.2%.

I believe the sixth home (if he came from a long way back and produced a finishing speed of 103.2%) would deserve a big upgrade as it seems very likely that the race was not run to suit. But if the seventh stumbled home with a finishing speed of 97.2% (also a 3% deviation from the median) having been prominent initially before fading, I don’t think the same big upgrade is deserved as he had every chance to run the kind of rating produced by the first three having gone through the race similar to them,

I will still be producing and displaying raw finishing speed % data for each horse in each race as these data are very valuable in their own right – they offer an excellent, easy-to-understand means of interpreting how a race has been run in terms of pace.


3: Notes for horses disadvantaged in some way in each race

This season I will be including very brief written notes for each horse that I believe has been significantly disadvantaged in each race.

The thinking there is simply that it is important to know whether certain horses in an upcoming race are likely to have been underrated from a speed ratings perspective in their most recent races.

This will be particularly valuable when looking at horses who have been disadvantaged recently but whose running style maps positively to the draw and conditions of the upcoming race. Without these anecdotal notes, their enhanced chances could easily be missed when looking at the speed figure data.