Daily Fantasy Sports, Draftkings/etc

In my hunt to try to get one those Brady NFTs had some money on DraftKings so played a few 1 to 3$ Premier league ones and it’s really fun. I like a lot better than normal fantasy because no big loss if make one bad lineup. I know it’s a -EV proposition with VIG and folks with high end simulation stuff, but trying to make my money last.

Ive found the optimizer type websites, but trying to find out a kinda basic question.

Historically for soccer has OPRK (opponent rank vs position) or FPPG (points per game) been the stronger predictor of outcomes assuming at least a minimum baseline “decent”threshold of FPPG. Seems like that would be a pretty basic easily known correlation factor but some quick googling didn’t show results.

I don’t get the Q and I don’t think you are approaching it correctly. I don’t play DFS soccer, but as any sport it’s all about lineup construction.

I think Awesemo.com has a promo for soccer now. You can also get a free two months RotoGrinders subscription if you deposit any amount to PrizePicks.com or MonekyKnightFight.com via their affiliation link.

If you do MLB/NFL/NBA dfs i can give you decent answers.

Opponents won’t change much in soccer. Valuable things to know are who takes corners and penalties

I know its not the winning factor, but anytime there is a variable there is a correlation and I’m just curious what it is.

The two variables prominently displayed for each player are OPRK (lower is better) and FPPG (higher is better). I’m curious how they correlate to actual points generated historically.

For example

player A has FPPG of 10 and playing OPRK of 8.

Player B had FPPG of 8 and playing OPRK of 12.

I think it would be pretty simple historical stats correlation to know historically which player would generate more points that day.

I already got all the basic stuff like set piece takers, wingbacks/etc. But opponent clearly matters for predicting potential for a multi goal game for example.

oprk are noisy and will give you more false data than anything useful.

just because those are two data point DK showing you don’t mean they have meaning. OPRK vs position is especially pointless because “position” is not a real life term, but a DK term.

1 Like

use vegas lines for that.

and again, lineup construction matters a lot more. you are not building a GPP team for a median outcome. Therefore stacking and ownership are far more important.

Yeah I figure given payouts weighted towards winning you want to maximize variance

It’s not just maximizing variance.

Baseball is an easy example. In a full slate there is literally 0% anyone will have the optimal lineup. You can go with 9 one-offs and hope all of them will hit home runs, but that will fail 99.infinite9% of the times.

What you should do is stack a team with 4-5 hitters and hope that team goes off. In most sports points are correlated.

In Football you correlate your QB with at least one receiver. Going with Aaron Rodgers as your QB and no GB receivers makes no sense, but it will show up if you just look at median projections.

Basketball is probably the only big sport where median projections kinda rule supreme because it has so little variance compared to other sports.

The way the Drsftkings EPL works you definitely stack because it’s only games kicking off near same time to each other so the pool of players is (generally) smallish so don’t want your forward scoring on your GK/etc.

Smaller slates dont make stacking more viable. Its actually the opposite in general

I meant it’s impossible not to because if there are like 2 games and your choosing 8 players there is overlap

Well first weekend done, cashed in 2/4. Obviously ran hot, but I think the entertainment value is pretty high for someone like me who enjoys soccer but doesn’t really have a specific team to care about.

The answer to your question is FPPG by a wide margin. Think about it this way: if you were building a statistical model, the outcome variable would most likely be FPPG. That conceptual model might look something like this:

FPPG = intercept + [predictor variables] + error

The OPRK is just one of potentially many predictor variables you might include. In addition to this, the problem with rank is it says nothing of magnitude which is what we really care about for winning fantasy contests. If the difference between ranks you’re comparing is relatively small in actual fantasy points then it’s very deceiving (and this is often the case in for fantasy sports). I never got into EPL but in the other sports that I played for years, an exceedingly large matchup differential might lead to a 20% predicted increase or decrease in predicted FPPG and often far less.

As well in the PL, the gap between the elite guys like bruno, kdb, salah etc vs random midtable dudes is huge. It’s like basketball. There’s a reason to own Giannis in a tough game over towns vs a bad one.

When I was writing solvers for this I used to include something called “fish optimal” which was the optimal lineup based on current FPPG in the Fanduel csv for lulz because it took an extra 0.2 seconds. I also did not want to run a lineup that looked anything like it in a tournament. However, I’ve seen several articles / simulations that show such a lineup is not trivial to beat, and I think this is likely correct as long as there are no obvious small sample shenanigans going on.


I was listening to a baseball dfs show today and the host was talking about their soccer product and he said “Premier League” in a way that made my ears bleed. I wonder if that’s how North Americans pronounce it.

Hockey and NFL are sports I’d take more matchup flyers. Hockey is basically random, matchups matter more in football. For epl/nba, you just see super high ownership of the elite guys for a reason. An average midtable striker might get like 8 goals/3 assists or so. An elite one will get double 2-2.5x that. It’s obviously taken into the price, but a strategy of a lot of mid tier guys usually doesn’t win.

Interesting, thanks.

I think I started thread out a little bit on the wrong foot. My questions really boils down to once your already looking at only the top or so available FPPG soccer players in each the DraftKings position then how much does opponent matter vs marginally higher FPPG. Seems pretty important, for example this week the specific example didn’t come up because they were playing on different days but I suspect (even before knowing results) Chelsea players vs Crystal Palace were dramatically better choices than Manchester City players vs Spurs. But curious where the line is when it’s less obvious.