Last time I wrote one of these, we were talking about over $15,000 in wins. Since then, I’ve added a couple of thousand in NBA winnings, but it’s not that I want to talk about today, but rather the start to Big Bash, and particularly our approach for the benefit of new subscribers. We’ve always been as transparent as we can, and I think it’s important to communicate when things are frustrating, as well as on the occasion of the the big wins. And, to that end, I wanted to give some quick thoughts on the first couple of Big Bash games, and using the cricket model to enter the guaranteed tournaments on Draftstars and Moneyball.
Firstly, the first couple of games have been pretty frustrating. I dropped about $550 across the two sites on the opening game personally, but I’m OK with that — given the overlays and the long-term record of the model, I’d make the same lineups and decisions again. I’ve written a lot on the site before about variance, but for those of you are new, here’s what my profit/loss chart looks like (to mid-November, so before the recent NBA wins):
The ROI isn’t actually as dissimilar last year to this year as that graph makes it seems, but the industry has grown significantly and the pools are much bigger. If we just look at cricket results from last season (using the same model), this is what we see:
In both cases you can see that there are losing streaks, sometimes extended ones. That’s variance — the model isn’t a time machine (if it was I’d be on a beach somewhere), it gives profitable long term projections, but in any given day or week the humans on the field are going to vary in their performance.
That said, our models are machine learning based, so always learning, and the humans that enter the lineups and make the decisions on how best to utilise that model can learn too, so I wanted to take a quick look at the first couple of games, and see what we can take away from them:
Thunder vs Sixers
My results: -$550 (best: 269 Draftstars, 626 Moneyball)
At one point in the second innings, with Henriques just getting started, I had a top 20 team at Draftstars. As it turned out, if you didn’t have Henriques and Hughes you were doomed. But how did the projections go otherwise? The below show the approximate difference between our projection and their final score (i.e. +1 means we projected 8, but they scored 9)
Haddin: Moneyball +1, Draftstars -1
Morgan: Moneyball -7, Draftstars -14
Patterson: Moneyball +1, Draftstars -3
Cummins: Moneyball -1, Draftstars +2
O’Keefe: Moneyball +16.5, Draftstars +37
Rohrer: Moneyball -3, Draftstars -10
Mennie: Moneyball +22, Draftstars +32
Roy: Moneyball -2.5, Draftstars -2
Sandhu: Moneyball +5, Draftstars -11
Henriques: Moneyball +44, Draftstars +98
Mckay: Moneyball -7, Draftstars -9
Green: Moneyball -25, Draftstars -47
Ahmed: Moneyball -6.5, Draftstars -30
Russell: Moneyball -15, Draftstars -25
Botha: Moneyball -13, Draftstars -21
Dwarshuis: Moneyball =0, Draftstars +1
Hughes: Moneyball +36, Draftstars +67
Bollinger: Moneyball +12, Draftstars +36
Doran: Moneyball -13, Draftstars -30
Abbott: Moneyball -17, Draftstars -51
Billings: Moneyball =0, Draftstars -4
Most of those players are well within their expected distribution (i.e. their projection +/- their standard deviation), but I waned to run through a few of them to give some insight into the projections, and I’ve highlighted in bold those worth talking about. The only one above that I thought last night I perhaps should have double checked to see if there was additional data we could add in is Hughes, who we had projected for a miserly score. Going into the match, Hughes had played in 17 T20 matches. He didn’t play in 15-16, and in 14-15 played in 7 matches, batting in 5, for a total of 100 runs. His Strike Rate was right around 100. Had I manually adjusted it, I can’t see I’d have given him much more than 10 points, so I think we just have to write it up to a career-best performance.
Henriques’ projection suffered as he was a (listed) all-rounder who announced in advance he wasn’t going to bowl. There was little in his batting record that suggested his projection should have been much higher than it was. Chris Green bowled tidily for little reward with only one wicket falling for the Stars, so again little regrets there.
That brings us to Mennie, O’Keefe and Abbott. With Mennie, two wickets from a bowler given just two overs is a much better performance than we expected (we did have him bowling less than the full 4, but given the result that’s a bit of a hollow victory). Abbott only got 3 overs, where the model had expected more, and for a bowler with a SR of 20, we had him taking about 1.2 wickets from a 4 over spell — he took none, and got hit for 8 an over. For O’Keefe, his 4 Economy Rate from 4 overs gave him sufficient bonuses and dot balls to significantly exceed his projection.
More importantly, the error rates above (when viewed as a % of the winning score to adjust for the DS scoring system) are well within the range from last year. They would have also led to a losing day last year, but I don’t think are anything to worry about long-term.
Heat vs Strikers
My results: -$18 (best: 104 Draftstars, 52 Moneyball)
I won’t do the full run-down here as this is already too long, and this game was nowhere near as bad — if you do this for a while you’ll have a *LOT* of breakeven days.
Dunk exceeded our expectations, but we were still high on him compared to his salary and he would have been in most subscribers teams, especially at Moneyball. The same largely applies to Weatherfield.
The big ‘fail’ from a projections perspective was Steketee, who we had on -1 but scored 49 at Moneyball (we had him on 35 at Draftstars where he scored 109, but again that 35 would have been enough to have him in a lot of lineups). Stekete has a SR of 17, and an economy rate of 9. Given the way we expected the game to unfold in Adelaide, we had projected him for slightly less than 1 wicket, and an economy rate of just over 10, but we hadn’t figured on the way the Heat used their bowlers. Stekete’s 2 wickets in the final over were worth 20 points, and his catch earned him an extra 5. Overall, no massive regrets on that one.
Alex Ross is probably the other (non-)play some would question, outperforming our Moneyball projection by 39 and our Draftstars one by 90. The short answer on Ross is that today was 1 short of his career best, and when you combine that with 4 6’s, and 6 4’s and a strike rate of 177 (vs 142.23 career), again you get a performance not many would predict based on recent and historic form.
On the upside, most other players were well within the expected range, and I saw a number of subscribers amongst the leaders; so I suspect some would have well outperformed my results.
Particularly in the early games, the model at the time it is run isn’t aware of the final batting order. This opens some opportunities, whether it’s cheap players like Weatherfield and Dunk who will be batting high in the order, or players listed at bowlers batting high enough that perhaps they should qualify as an all-rounder, there are opportunities to exploit.
As always, we emphasise that Fantasy Insider is a research tool. We don’t sell lineups, and we don’t recommend users just enter our default team. Much like with the NBA, there are some adjustments you’ll want to make when entering lineups; whether that’s stacking (e.g. opening bowlers from team A with mid-order to lower-order batsman from team B), locking in players (e.g. Weatherfield and Dunk) to a set of lineups, or taking contrarian approaches (I didn’t do this, but entering Henriques based lineups in the first game would classify, given most users were avoiding him as a non-bowling all rounder).
The model and projections will continue to improve as the tournament progresses, and I still have full confidence that come the end of the tournament, the projections accuracy will be as good if not better as it was last year, where I ended with a 52.5% ROI. The ROI will almost certainly be down (bigger tournaments, better competition), but I would expect the profit to be as good if not better.
You can sign up for our Big Bash tournament pass here. Good luck the rest of the way!