Byron Rogers was on the podcast in 2014 and it was refreshing to discuss his efforts in quantifying both breeding and type when identifying yearlings to purchase. That discussion is well worth a listen, but if you want the super-simplified summary of how he tries to find Group performers: (1) Young-ish sire and dam (2) Fast dam (3) Good heart, body type and DNA (4) Raised on a suitable farm (5) Sent to an excellent trainer
Byron is back this week to talk about all that’s happened since then.
- Why breeding a stayer is so difficult
- Why non-stayers can win the Oaks
- How the trainer effect is often under-estimated
- Why their success tends to be at the ‘edges’ of the horse population
- How and why the model gets smarter every 6 months
- The machine learning software he (and the NY Yankees) use
- How some of this can be applied to the betting market
Today’s Guest: Byron Rogers
Dave Duffield: I was chatting to you last week to come on the show, but you had a nice weekend.
Byron Rogers: Yeah we had the filly Indarra for Phil Sly we’d purchased as a yearling out in New Zealand at Karaka for 150 thousand and she broke her maiden in good style in a stakes race which is obviously the best way you want to do it with a Filly.
Dave Duffield: And so, we’ll speak in general terms as we chat but specifically for her, what appealed when you’re going through the yearling catalogue and then also doing the various tests and inspections that you do.
Byron Rogers: I think the main thing with her was, she’s by Stravinsky who obviously hasn’t had a lot of good horses recently, he’s been a very good sire but he hasn’t in more recent times hasn’t had a lot of good runners but when you physically looked at her and Phil Sly was pretty specific about wanting to get fillies that were going to run to ten furlongs. He likes horses that can try and get a mile to ten furlongs and you know he has got a lot of luck with horses like Mosheen and those sorts of horses.
So he was very specific about what he was trying to get to. When we looked at her physically she didn’t look like a one dimensional horse. She was very much like her broodmare sire, like a Zabeel and you probably see that when she’s at the races you know she does have that Zabeel attitude to her. But physically she’s a well-balanced filly and we just thought she fit the bill and when… Obviously when we do what we do in terms of looking at the genetics, DNA markers in relation to muscle fibre type she had the right sort of balance for what we see to be a horse that would get the ten furlongs. More importantly she also had the cardiovascular capacity. She has got a good sized cardio and so a Filly like her, for us, stood out. We did have to pay 150 thousand which was a decent price for a Stravinski filly at the time and she has worked out.
I mean she was a bit unlucky, we probably thought she was going to be a Filly that would come to the races and be competitive in a VRC Oaks because what we find is that an early in their 3 year old year the genuine stayers are actually a little bit behind the horses that aren’t genuine stayers in terms of…. You can win an Oaks with a filly that actually doesn’t really sort of stay a mile and a half given a VRC Oaks. They can be milers that get there by the fact that they’re more mature and more forward and we thought she was a filly like that but unfortunately she got an ulcer under her tongue we sort of had to back off her.
It was a bit of a blessing in disguise we sort of got her going at the tail end but she’s a filly that will end up getting to ten furlongs whether she gets to the mile and a half in the Autumn remains to be seen. The other horses tend to catch up to you there so that’s the new challenge, just to see how Robert Smerdon handles it.
Dave Duffield: So Indarra is one example, but just going back to the last time you were on the show, to summaris in layman’s terms I’d say what you guys do is you are trying to quantify both breeding and type but the recipe is often young parents, fast mum, good heart and body type, the right DNA, raised on a suitable farm…and you’re trying to buy at the right price and send it to an excellent trainer.
Byron Rogers: You’ve simplified the complex there very easily!
Dave Duffield: Well our last discussion was quite complex, I really enjoyed it.
Byron Rogers: Yeah, I think the main thing you’ve hit on there which is true is the trainer effect is quite large. One of the things I’ve said to a number of people we work with… You know they say I’ve got this horse and I ask who’s it going to be trained by? People underestimate… That’s the first thing I think that people underestimate. What you do, in terms of, who you give the horse to actually has a very large impact on the outcome. I think the trainer effect is underestimated by a lot of people.
I think you’ve actually… You are in the industry and want to be successful there’s a reason why the really successful trainers have a lot of horses in work because they can, not just extract the best out of all horses but they can certainly maximise the really good horses and that’s what you want to be able to do. But I think that a lot of the simple things we look for… In some ways Indarra was a little bit of a reverse of what we look for in that Stravinskiy was an older stallion and this was a slightly older mare.
Again, when we get in there and we start to look at the DNA markers and say okay, you know, what is this horse in terms of its genotype and does what we look for in terms of cardiovascular capacity match up with that? One of the things we see a lot of is horses with sprinting cardios that are distance horses or vice-versa. So we see a lot of horses that say, okay this horse should be 6-8 furlong horse and its got a cardio like a two mile jumper or its the other way around. We see a lot of horses that are supposed to be distance horses but just don’t have the cardiac capacity to get that far in high competition. So we’re marrying that all together in one database. We’re up to 7 and a half thousand horses now so we’ve got quite a large database with a lot of outcomes.
We’re able to see those things fairly quickly I will say that we’re probably getting better as we go along at the edges of the population. So if you sort of look at the population and said OK well precocious early 2 year old sprinters that are on one end and later maturing 4 year old distance horses are on the other well we’re actually very good on those two edges. Where the middle part is… Which is those horses that are three year olds that might run as… to ten furlongs or something. They’re not as easy as the edges and I think the reason for that is, we can find horse where they’ve got real physiological advantage.
So if you’re talking about an early two year old like Fontiton and we had a filly Motown Lil running the fourth in the stakes race on Saturday as well and she was a bit unlucky in that race. But they have genetic markers which sort of say that they’re going to be sprinters but they’ve also got the advantage of a very fast, early maturing cardio which is easy to train and that gives them at particular pockets of time earlier on in their career, they get to the track easier, they get to train harder and they can respond to training better.
So we’re very good at finding those horses with various points of physiological advantage. On the flip-side of that you’ve got sort of a horse like Keen Ice, who up here in America I worked with the guys who bought him and he was one of the few horses that has beaten American Pharaoh, he beat him in the Travers but if he’s in Australia he’d be a two mile horse, he’s a definite stayer so, you’ve got these ends of the population which we think are very good at and as we go along we keep seeming to learn.
Dave Duffield: And so, with the edges of the population as you talk about it, it probably has some betting applications in that you guys know what your strengths are and can utilise that. It’s probably similar from a punting perspective – people should know where they’re edge lies and then be able to maximize it.
Byron Rogers: Yeah, again it’s no different to as you say a handicapper or punter, my brother for a while was playing professional poker and they know what hands they play well and what they don’t play well. I’m sure there are punters listening to your show here that know that they’re very good at trifecta handicapping or they know that they’re very good at different bet types and under different bet conditions i.e. they know that they’re no good in the wet or whatever. It’s no different. We just have a lot of data and we use machine learning languages, we use a program which Microsoft use which is in Xbox it is in Bing, the New York Yankees use it for their analytics.
We use that program to, when we update results so, like if we had Fontiton she was an unraced yearling. Now that we know what she is, the database learns off her and says okay well I know, you know, it basically confirms she rated well for us and confirms its rating because we’ve updated her results so, yes she was a superior runner. Each time we do that, we learn a little bit more and the database starts to get tighter and tighter and we start to sort of be really able to find horses and work out why they’re good horses and I think you know that’s something that as a punter you should be doing that anyway you should be learning too. Why you make mistakes and where you make them and try not to make them again.
Dave Duffield: So, you mentioned the machine learning from the Yankees, Bing, yourself, I’m sure I’ll be asked this questions so what is the software?
Byron Rogers: It’s a program called Azure ML (machine learning) you can go online and look it up. It’s free to use to a certain level and its got a whole lot of different modules you can stick on there and test out your data and it’s very much plug and play and actually the good thing from my viewpoint is that it puts out an automated web service. So if I want to say, get a whole lot of data and just ping it up against the web service and automatically find out answers, well that’s what it does.
Anytime we get a horse that we test, I literally put all the data in press calculate results and it runs the data points up against the API and it returns back the score that, which is based on, you know, all those previous 7500 horses it comes back with a score instantaneously. There’s a whole lot of other machinelearning options like Rapid Midner and Big ML that you could use but that’s the one we found to be the best for our means and it’s certainly helping us get closer and closer to what we would like to be buying.
Dave Duffield: So every 6 months you’re feeding the new data in and re-testing your original hypothesis or assumptions. So basically you’re expecting to be more and more successful over time.
Byron Rogers: Yeah, every year we bought 12 horses in that first crop for Matchem Racing and two of them are already stakes winners so that’s one out of 6. We think there’s probably another two there that are probably stakes level and whether they get to that or not… Even if only one of them did that would be three out of 12. When you consider that the industry average is about one out of ten being… And that’s a good result, one out of ten being a stakes winner, to get three out of twelve if it happens that way is a very good result but I think going forward, we were underbidder on Capitalist as a yearling and, you know, James Harron bought the horse and he, he came up straight afterwards and said I thought I paid too much for a Written Tycoon then I saw you guys were the underbidders on it.
We’re happy with where the program is sort of helping us to find these horses we still do a lot of work on the physical side of stuff with Merrick Staunton and Robert obviously looking at a lot of horses and we’re doing some work with David Hayes and Jason out there at Lindsay Park so there’s still a fair bit of horsemanship to it. You don’t completely rely on the data but what the data can do is, tell you where you’re not to go and where you should really be strong and when we find one that we really want, we strap on and get as much capital behind us to buy the horse.
Dave Duffield: And, has a lot of that recent analysis or updates been focused on sprinters?
Byron Rogers: Yeah, its been one of those things where we’ve done the most recent research we’ve done is to do with variation in muscle fibre types. We’ve had a look at alot of differences between elite sprinters and non-elite sprinters. Sort of the original work that us and other companies to do genetic testing on was on a chip that had 70 thousand markers across the horse genome. Well just earlier this year they released one with 670 thousand markers so we’ve used that to specifically look at sprinters and say OK well if you take the group of horses that are elite sprinters what is the specific differences between them as a group.?
I think the reason we did that mainly was because about 50% of the population in Australia, in terms of yearling population, commercial yearling population ends up being sprinters in genetic terms. They’re very high on type two muscle fibre so we basically said well look, if half the population there in Australia is sprinters and this is a totally different conversation we could have as to why that means we can’t produce decent stayers in Australia but if half the population is sprinters then we really need to know everything we can possibly know about sprinters. We know why certain sprinters will have different cardio types and what that means but we tie that into the genetics and some biomechanical features and we get a holistic view of it all and we get a much closer and richer view of what actually takes to be a good sprinter.
We’ve just literally finished that research in the last month and that’ll be something that we’re using at the Magic Millions and going forward with clients that we’re going to be dealing with over the yearling sales series. Only because from our viewpoint that’s where the money is. If you’ve got a Golden Slipper winner or you’ve got a horse that wins a good race down the straight at Flemington or one of those races you’ve got a very much, an elite horse in the viewpoint of many breeders and owners in Australia. We just took that view that that was an important aspect for us to focus on.
We really worked hard at working out what makes a good sprinter. The other flip-side, the reason for that that is also, even though we’re very good… I think, very good at long distance stayers I think the further you go out in distance the harder it gets in some ways until you get to that part where they’re really long distance horses. The reason that is, for one example, the horses, say when they’re galloping they have a one to one stride to inspiration.
So when they’re breeding, well their stride and their breath is the same so with a… As the horses get out in distance you want further and further for them to have longer strides and that sort of stuff and having enough speed. But the problem there is if you’ve got a slight biomechanical inefficiency or something that, you know, you don’t have quite as strong a cardio, everything can go wrong and it compounds in every stride. So we just found that sprinters and sprinter miles are much easier in terms of being able to sit down and reference out as to why certain genotypes…
What they need to be good horses with certain genotypes. Now we’ve got a database getting in the thousands we can start to really see where populations fit and get a very clear idea as to what makes them tick.
Dave Duffield: So, you mentioned earlier about some horses from the work you do can be strong, certain aspects as a sprinter and then other indications suggest they’ll be a stayer. Do they end up as milers or neither or slow or…
Byron Rogers: It’s a complex thing to make simple, but if you miss the target on the distance horse. Like, if you’ve got a horse that genetically has a higher proportion of type 1 muscle fibre as opposed to the type 2 sprinting muscle (so they’ve got very much oxidative distance muscle fiber) with those horses if they just have a slight inefficiency in one thing, so, if their cardio is not quite big enough, not quite strong enough, if they’re leg length isn’t quite the right way and they’ve sort of having to lift their knees rather than stretching out from their ankles, you end up with a very slow horse basically because of the fact that there is a compounding effect.
But when you get them right like we did, say with Keen Ice up here in America, they’re very good, they’re Group 1 level horses. I think that breeding a stayer is very difficult and it’s just as hard to buy a stayer and people that go to the yearling sale and say, I’m going to buy a stayer, I think that’s a very admirable attempt but I think it’s really hard to do consistently like a guy. Bart Cummings who could come and buy stayers, a lot of stayers, but I think it was more not just his eye for looking for a horse but more that he could train them. More that he could really train a stayer well rather than being able to consistently buy them. Stayers are very difficult to find but when you do find them they… The good ones stand out a little bit more than the average.
Dave Duffield: So, you talk about Bart Cummings or trainers in general because you did mention the importance of the trainer effect. How much does that vary depending on the type of horse you’re sending them too. So, would you have a different shortlist of trainers for a sprinter than you would for a stayer?
Byron Rogers: Every trainer likes to think they can train every horse well, I think there are only a few trainers that can actually do that. Like obviously, we have horses with Robert Smerdon and I think Robert’s’ a trainer that trains horses that win Blue Diamonds and he can train horses to win steeplechase races and the Sydney Cup and all sorts of things like that. So he can train the full spectrum. But there are very few trainers like that. I think that the vast percentage of trainers, the way they’re trained in Australia in terms of the ‘5 home 2’ or the ‘4 home 2’ type training where you’re just letting them squeeze up the last two furlongs…that lends itself to sprint genotypes so I think that a lot of the trainers will find that’s why there’s probably 50 per-cent of the population in Australia have got sprint genotypes because that’s what generally…
It might be a bit of a chicken and egg situation but that’s generally what trainers find easier to train in Australia because of the way that they train them and where they train them. I think that really good trainers can adapt. You’d sit there and say, well Darren Weir hasn’t had an early 2yo and then he comes out with an early 2yo this year, he’s obviously thinking about how he can adapt his training and come to that. For the average trainer, who hasn’t had the experience either producing a stayer or a classic horse. I think it’s that much easier for them to train a sprinter and I think that you can find those horses a lot easier for them to win good races with.
Dave Duffield: Excellent, well we’ll leave it there for now Byron. Now guests on the show are a bit like my kids, I’m not allowed to have a favourite, but what I will say is I learn something each time you come on the show so I appreciate your time and we will definitely check in with you again in 2016.
Byron Rogers: All right, thanks Dave, keep well.
Get More Betting 360
Make sure you don’t miss our punting tips to come! Subscribe on:
or you can directly download this episode by right-clicking, Save As Here.