The Rotoquant

Describing and predicting the world of sports in the language of mathematics

Month: March, 2014

Valero Texas Open Picks, plus a look back at The Arnold Palmer Invitational

No time for real analysis this week. Last week was a good week, hopefully you didn’t get burned by Bubba (who was projected outside the top 5 here, so perhaps that saved you). Gotta run but the charts and the picks are below! One quick comment on last week’s performance: I need to look into the strange phenomenon of rank-order (%cut made, %top 10’s, etc.) vs RMSE performance and see if I can’t figure out what’s happening between the Machine and Notorious from Rotogrinders.com. I’m starting to believe there’s a non-linear method of combining these picks (or even, say, a Top 25 classifier and making this whole thing a classification problem) which would take their predictive power to a new level. Stay tuned!

Look back at The Arnold Palmer Championship

The same caveats as always apply to this analysis.  

As usual the comparison picks are from Notorious at Rotogrinders.com, a very well respected name in the daily fantasy sports industry. If you don’t know what these graphs are showing (and you care) then just go to first fantasy golf post that’s hyperlinked above – it contains all the explanations of the relevant concepts used in this analysis. 

Below is a grid showing Notorious’ draft street points projections (DSP), the rescaled draft street points projections (RescaledDSP), the machine’s RescaleDSP and the actual RescaleDSP’s:

Picks_Arnold Palmer Invitational_2014_MFP_Grid

Looking at the RMSE, this week the Machine did extraordinarily well when compared to Notorious on this measure (and you can think of RMSE as a measure of the ‘general goodness’ of the predictions). The discrepancy between this RMSE measure and the %Cut, %TopN figures is the ‘Rank Order vs Absolute Performance’ issue I discussed above. It’s happened ever since the switch to rescaled figures (as opposed to % mean field numbers), and it bears investigation (hopefully before the Masters, but no promises):

Picks_Arnold Palmer Invitational_2014_RMSE_Grid

Again, almost no blending with Notorious’ methods produced the optimal RMSE last week:

Picks_Arnold Palmer Invitational_2014_Blending_Plot

The farther left you are on this line, the more of the machine’s projections you are using, the farther right here represents Notorious. So last week was a rare example where the Machine outperformed Notorious on average such that blending the two predictions would not provide any average lift in the resulting RMSE. But again, like last week, that’s a very loaded statement.

Here’s the cut percentage for both of us:

Picks_Arnold Palmer Invitational_2014_Made_Cut_Grid

As well as the machine did at RMSE reduction, it was ousted by Notorious in the percentage of golfers correctly projected to make the cut. So the interpretation here is that Notorious’ method did a better job of predicting the rank order in terms of fantasy performance, while the Machine did a better job at predicting the actual performances.

Finally for those less mathematically inclined an idea of how many of the ‘top-n’ players were predicted correctly in each method.  Keep in mind all the caveats around who gets included in this analysis… If I get the comment ‘But, so-and-so didn’t finish in the top 10!’ I’m going to be very disappointed, dear reader.

Picks_Arnold Palmer Invitational_2014_TopN_Grid

I’d say we both had an average to below average predictive outing based on this chart, but it’d only be an artifact of the data feed not having been updated this yet week (my apologies). From my perspective, last week was quite solid – I even picked myself a 50-1 winner in Matt Every on Bovada.com where – as an experiment – I’ve been placing 5-7 $1 bets to win, using the Machine’s picks as a starting point, for the last few weeks … felt pretty good to pick a first-time tour winner for the first correct winner pick of the season – certainly making the experiment profitable for now.

At this point everything is leading up the to Masters, where DraftStreet has already announced the first ever $100K guaranteed tournament for fantasy golf.  DraftKings answered the call with their $300K Spring Golf Classic, maintaining their spot in my mind as the premier golf offering in the DFS space.

There are indeed great things to come, but right now we’ve got ourselves an Arnold Palmer Invitational to dominate!

Picks for the Valero Texas Open

Here they are:

Picks_Valero Texas Open_2014

The red line here represents the cut, based off of these projections (the cut being returned to its normal 70 count).

Good luck everyone in your contests.  Talk to you next week.

Arnold Palmer Invitational Picks, plus a look back at The Valspar Championship

Here we go… Tomorrow is probably one of the most exciting Thursday’s of the year – the Arnold Palmer and the NCAA tournament both kick off mañana. The Valspar Championship was fun, and I obviously love watching golf but there’s just something special about the NCAA.  For those of you who love computation and are participating in a March Madness bracket, I found this great site for creating small programs to fill out one’s bracket computationally: www.codersbracket.com. I’ll be entering the bracket created by my 20 line little program in the Buffet Bracketology Billion-dollar Blowout (name not trademarked), can’t wait to see how it goes. 

Look back at Valspar Championship

The same caveats as always apply to this analysis.  

As usual the comparison picks are from Notorious at Rotogrinders.com, a very well respected name in the daily fantasy sports industry. If you don’t know what these graphs are showing (and you care) then just go to first fantasy golf post that’s hyperlinked above – it contains all the explanations of the relevant concepts used in this analysis. 

Below is a grid showing Notorious’ draft street points projections (DSP), the rescaled draft street points projections (RescaledDSP), the machine’s RescaleDSP and the actual RescaleDSP’s:

Picks_Valspar Championship_2014_MFP_Grid

Last week was somewhat indifferent from a prediction standpoint. The average prediction was actually quite good – but I just feel like the myth of parody is rearing its ugly head. If you picked any of the last four winners on tour, well, you’re a liar! John Sendon? Really?! Where’s Adam Scott, Dufner, Luke Donald, Justin Rose? Wasn’t there some guy named after a jungle-animal that was pretty good at one point? I mean, the household names are really falling flat thusfar this year, but I don’t think the right thing to do is to go against the statistics and start picking no-names. In fact, I’d argue that money has probably shifted hands from sharps to fish in this first part of the PGA season – which bodes well for the back half of the season assuming the world becomes more rational again.  The point is that even a highly diversified sharp would probably not have a squad with John Senden on it last week, or even Patrick Reed the week prior week given the strength of the field at the Cadillac.  Of course – and according to one of my favorite adages in finance – the market can always stay insane longer than you can stay solvent. But I don’t recommend jumping ship because you haven’t picked a winner this year; I don’t buy the parody argument (yet), and I look forward to the best golfers re-establishing themselves moving forward. Hopefully it starts with Scott this week!

Anyway – on with the analysis… Looking at the RMSE, this week the Machine did extraordinarily well when compared to Notorious on this measure (and you can think of RMSE as a measure of the ‘general goodness’ of the predictions):

Picks_Valspar Championship_2014_RMSE_Grid

It turns out that no blending with Notorious’ methods produced the optimal RMSE last week – but there’s a valuable lesson to be learned hiding in the details of this statement. Check out the blending curve below:

Picks_Valspar Championship_2014_Blending_Plot

The farther left you are on this line, the more of the machine’s projections you are using, the farther right here represents Notorious. So last week was a rare example where the Machine outperformed Notorious on average such that blending the two predictions would not provide any average lift in the resulting RMSE. But that’s a very loaded statement.

Fantasy sports is not always about making the best average prediction.  It’s often times enough to have 4-6 excellent predictions: if you project 6 golfers perfectly as finishing in positions 1-6, and those are the golfers that make up your team, it really doesn’t matter how well you predict the other 144. The overarching theme here is that more accurate average predictions do not necessarily translate to better fantasy performance in a world where total diversification is impossible. Unfortunately for people like myself I don’t get to pit my predictions for every single golfer against every single other persons predictions for every single golfer. I make the best predictions I can, and then pick a group of 6-12 from different points in the salary range that I’ve predicted to do well (alas, John Senden didn’t make the cut last week) and try to setup some semblance of diversified rosters using that group. Did you catch the absurdity there? I have what I consider to be state of the art projections for 144 golfers and but only 12 of them really matter to me in a given week… yikes… it’s an open invitation for increased variance – but there’s only so much time in the day and, to be honest, trying to follow the performance of more than 12 golfers during a tournament gets annoying.  

Here’s the cut percentage for both of us:

Picks_Valspar Championship_2014_Made_Cut_Grid

As well as the machine did at RMSE reduction, it was ousted by Notorious in the percentage of golfers correctly projected to make the cut. So the interpretation here is that Notorious’ method did a better job of predicting the rank order in terms of fantasy performance, while the Machine did a better job at predicting the actual performances. 

Finally for those less mathematically inclined an idea of how many of the ‘top-n’ players were predicted correctly in each method.  Keep in mind all the caveats around who gets included in this analysis… If I get the comment ‘But, so-and-so didn’t finish in the top 10!’ I’m going to be very disappointed, dear reader. 

Picks_Valspar Championship_2014_TopN_Grid

I’d say we both had an average to below average predictive outing.  Predicting 1 of 5 top fives is nice – but only predicting 1 of 10 top 10’s is average at best.  Notorious bested the Machine in picking the top 25, but I wouldn’t say either of us is too excited with 50% or less in this category. 

Ok, glad to be back on track with the Notorious comparisons, and the more detailed analysis of the projections. The next interesting thing I’d want to explore is how the projections are different in specific cases; while on average the Machine does well, it might be the case that the bulk of its errors are, for instance, related to players on the Champions or Web.com tour who are ‘playing up’ on the PGA. Or, perhaps, the Machine is going wrong wherever the data is very thin about a golfer – giving them too much or too little credit. Unfortunately these types of investigations will have to wait for another day (my next vacation!).

At this point everything is leading up the to Masters, where DraftStreet has already announced the first ever $100K guaranteed tournament for fantasy golf (they are running satellites into this tourney this week for the  Arnold Palmer Invitational).  DraftKings answered the call with their $300K Spring Golf Classic, maintaining their spot in my mind as the premier golf offering in the DFS space.

There are indeed great things to come, but right now we’ve got ourselves an Arnold Palmer Invitational to dominate!

Picks for the Arnold Palmer Invitational

Here they are:

Picks_Arnold Palmer Invitational_2014

The red line here represents the cut, based off of these projections (the cut being returned to its normal 70 count). 

I’d be remise if I failed to mention that the top projected golfer here as withdrawn from tournament; if you have Tiger on your squad tomorrow you absolutely cannot point the finger my way 🙂 (and God help you, he’s expensive). 

Other notables – Scott and Mahan I always like, other notables include Patrick Reed (who the machine seems to be taking quite seriously after the win), as well as Will-Mac, Molinari, and Every. 

At any rate – time is still short, but I’m happy to get back on track with the blog and I apologize for the delays during the transition from %MFDSP to RescaleDSP – those bugs were easier to fix than I imagined but 15 minutes can be hard to scrounge up these days. Good luck everyone in your contests.  Talk to you next week.

Valspar Championship Picks

I wanted to write a post detailing how last week’s Cadillac event (and to some extent the Puerto Rico Open) causes systematic biases in pricing. On the one hand, almost every stats provider I know will list Kiridech Aphibarnrat as having made the cut, even though he finished T67 and a gut wrenching +26 over par. And Chesson Hadley has the distinct honor of winning out in Puerto Rico over a field lacking the best 50 golfers in the world – well done sir. 

I had aspirations of  detailing how these data processing miscues (nobody ‘made the cut’ at the Cadillac, and winning in PR is not the same thing as winning elsewhere) provide an opportunity for all of us data hawks to make a killing in this coming week. I know that the data providers, the data from which all the major DFS sites make their salaries, is not nuanced enough to capture the relevant facts about last weeks tournaments.

perhaps you can conclude some important thoughts of your own from the above two paragraphs… But life is short, I don’t have time to fully convey my thoughts, so the most you’ll get out of me this week are the projections. So without further ado, let’s get to it.

Picks for the Valspar Championship 

Here they are:

Picks_Valspar Championship_2014

Red line marks the cut – best of luck to you all! Talk to you next week where I can (hopefully) reinstate Notorious’s comparison – but I can’t imagine work letting up for some time.

WGC-Cadillac Championship Picks

So… The other week I wrote a post about variance vs diversification in DFS. The point of it was to make sure we used the word correctly: variance being the expected week-to-week variation in performance, diversification being a risk-management concern about the size of one’s exposure to a given player, opponent, or game-type in a given week. Variance is out of your control (but should be planned for with proper bankroll management) while diversification is your responsibility.

Last week, variance  struck… hard… 

Two of my top four projected golfers withdrew from the tournament (Tiger and Matsuyama). One more withdrew if you look at my top 10 (Rose, though at least he did it early enough that you could sub someone else in for him). Schwartzel, someone who was on everybody’s short list of easy picks last week, missed the cut by 7 strokes.  Henrik Stenson, the 3rd ranked golfer in the world, also missed the cut. It was DFS Black Tuesday of the highest order – a several sigma confluence of events that blew my precious math up, right in my face. I was slack-jawed going into the weekend as every roster I put up ended with just three golfers still in play. I had no chance… so I went and played golf with my dad, and did some real work and was better for it.

But this stuff happens! It’s sports, it’s reality – as much as you put math to it to try and make it predictable it’s not math. There it is, that’s variance – that thing that takes a regularly profitable player and smacks him in the face with losses across the board. Were my bankroll practices perfect? Probably not – I admit to falling in love with a roster every now and then, and with a focus on weekly DFS like PGA and NFL it’s tough to wait all week for an event then only put up 10% of your bankroll on the weekend. But I certainly wasn’t all in, precisely because DFS Black Tuesday can strike any time. With solid bankroll management you will always live to fight another day.

And it’s a good thing too – because this week we have the World Golf Championships series Cadillac Championship event at Doral. There’s a bunch of stuff all over the internet about the face-lift the course received since last year, and the strange format of the tourny (no cut, 69 golfers). In a perfect world we might tone down our treatment of course-history to account for these facts, or do a case-study on how newly revised courses have played out in the past, or the effects of a no-cut tournament on decisions like whether you should go ‘stars and scrubs’ or take a more middling approach… But this world isn’t perfect; I hate to blog-and-dash but I got ‘thangs to do and too little time to do ’em! Suffice to say the course is harder, Tiger Woods is banished from my lineups forever (so everyone place your bets now on a sure Tiger victory), I don’t trust Rose, I don’t trust Sneds, and I don’t trust Phil (how could you, just look at him)… But I’ve left Phil in because, well, he’s always helping the kids and he’s got such a creepy vibe that I wasn’t sure how he’d react if he knew I didn’t rank him this week… Seriously though – what is this?

So, without further ado, let’s get into it.

Picks for the Cadillac Championship 

Here they are:

Picks_WGC-Cadillac Championship_2014

Obviously there’s no red line here, since there’s no cut. And, as I mentioned, I’ve removed the players I no longer have eyes for (but all of them, Rose, Snedeker and Woods were projected quite high. Woods was projected to win… but barely. If you have the cahones, by all means roster him. I think he’s having a press conference Wednesday to discuss his back, for whatever that’s worth. Maybe he will do us all a favor and withdraw early this time).

I think Adam Scott is a solid choice to win. Steve Stricker is also an absolute stud IMO. Fair warning – age is not taken into consideration here and Steve Stricker could be my dad, if my handicap were 14 strokes lower and I was hilarious (obviously not the case). That being said, maybe, that projection is a bit overzealous. Then again, the average winner at Doral tends to be an older player, so who knows.

As for players to fade? I’m amazed at how low Dustin Johnson is projected – I could see him embarrassing the machine with a very solid tournament. Victor Dubuisson is also an interesting fade considering his recent form… Also low on the list are Mahan, Stenson, Westwood and Schwartzel – some pretty big names. But this is one of the best fields all  year, and somebody has to finish outside the top 25.

One final note about the happenings in the DFS community – this week is DraftKings.com largest PGA tournament, a $75,000 guaranteed event, should be fun. Also – DraftStreet.com announced a change in how they score golf, meant to reaffirm that each round is an independent event (according to law, one criterion that must be met for DFS to be legal is that it has to span multiple real-world events).

Good luck everyone in your contests.  Let’s hope variance swings the other way this weekend. I also hope to get the Notorious comparison up and working over the weekend (time permitting) so next week’s post might be more colorful. Talk to you next week.