Who wins a T20 game?

Theories are going around for what is the best predictor for who will win a T20 game. Some look at the number of sixes hit. There is also the thought that if you score in more overs at above nine runs and in less overs at under five than your opposition, you are far more likely to win. But from watching all the T20 I have this summer, I started to wonder if teams making basic errors, either forced or unforced, might be the best predictor.

My system was profoundly flawed, subjective, and largely based on my personality. But regardless of that, I carried on. An error was a ball that wasn't to plan. For bowlers, full tosses, wides/no-balls, half volleys, short and wide, too straight without cover, and fielding errors. With the bat, it was not capitalising on one of those errors, and plays and misses. A wicket could be an error for one team if the ball was bad but the shot was justified, and some wickets, like Glenn Maxwell's from a Shane Watson full toss in the Sydney Thunder-Melbourne Stars match on Wednesday. The overall tally for the errors was 48 for the Thunder and 35 for the Stars.

I chose this game to look for errors because these two sides have only one win each in the tournament. I knew that the Thunder were batting poorly (when Cummins was your top-scorer at No. 6/7 until Wednesday, your season has gone horribly wrong), their bowling was ordinary, and their fielding has been about as bad as professionals can manage. The Stars might have won one and lost one, but their win came about from top-class batting after a pretty ordinary bowling performance. And in their loss, they were destroyed.

So far in this tournament, 71% of their runs have come from their top three. And that is despite Rob Quiney having sat out the last two matches with an injury, following a brutal first knock. No other team, not even the notoriously top-heavy Brisbane Heat, has relied on their top order more than the Stars. The Stars are dependent on their top-order by 15-16 percentage points more than most. Not to mention that the reliance on the top-three has gone up 17 percentage points compared to last year. When that top-order was up against a Thunder bowling line-up that is the second-worst in the league - not because of talent, but some poor deliveries their bowlers sent down - combined with the worst fielding side, it was carnage early on.

But the Stars' top order pushed hard, and the Thunder kept taking wickets. When the Stars lost their fourth wicket, it brought together Marcus Stoinis and James Faulkner. Their partnership featured one boundary in ten balls. Faulkner, known as the finisher for his antics in a handful of amazing chases, is not a finisher in T20, he's more of an accumulator. Batting at No. 5/6, he scores more than most players, but at a strike-rate of 126/121 depending on the position, which is about average for the Big Bash. Faulkner managed 12 from 15; Stoinis, 12 from 12. Instead of fireworks from a decent platform, both players struggled to hit boundaries, and the Stars finished with a lower score than they should have. They scored 20 for the loss of five wickets in their last five.

With data provided by Krishna Tunga of All that cricket, you can see that the Stars are, by far, the worst bowling side when it comes to the number of wickets and runs per over. So when you make 166, after all but declaring your last five overs, you have to take wickets. And here they were lucky as they ran into the side with the weakest top order.

The Thunder's top three have made around 30 percent of their runs so far this season, by far the lowest of any team over the last two seasons. Tonight, they scored 36 off 32 balls. The Thunder have lost Jacques Kallis, Usman Khawaja and Michael Hussey, who combined to score 788 runs last season. Tonight, they also had to contend with the absence of Andrew Russell, though his form has been so bad you wonder if his hamstring went on strike and wasn't injured. But it still meant that Pat Cummins was batting at six.

The Stars also only needed one more wicket, and the game was probably theirs. But they haven't taken many wickets this year, their best two bowlers are Beer, who contains, and Zampa, who can take wickets but mostly slides the ball through looking for dot balls. What the Stars needed were errors from the Thunder.

The problem was Eoin Morgan. While his innings may have looked bizarre at first glance, Morgan's career strike-rate in the Big Bash is 16 lower when chasing. Like Dhoni and Bevan, he knows that most of chasing is making sure you are set, and that is completely the case in the low and middling chases. Had Morgan made more errors in the last five overs - he made barely any, according to my "data" - the Stars would have won.

Perhaps in a better time, Faulkner would have bowled the slog overs. But he isn't the bowler he once was, and he hasn't bowled much in the slog overs this year. His slower balls have been worked out, and that makes his faster balls less important. Had John Hastings been fit, he also would have bowled in the slog overs, where he is excellent. So the bowler of choice was Ben Hilfenhaus, and cricket analyst Ian DW tweeted that according to his data, Hilfenhaus is not a very good death bowler.

So it was the cool-headed chaser up against the fast bowler with the average record at the death. For one over, when Hilfenhaus could ruin Pat Cummins' timing with slower balls, he looked good at the death.

During the entire game, the key winning metrics showed that both teams were equally good, or bad. The Thunder outscored the Stars in terms of sixes, but the Stars had more boundaries in total. The Thunder had one extra over of nine runs or more, and both had the same number of overs of five or fewer runs. The fielding errors were pretty much identical for both.

While my error counting system showed that the Thunder had 13 more errors than the Stars, I also divided them up into forced and unforced. When the Stars batted, they made 11 unforced errors, which certainly helped a struggling bowling and fielding side. When the Thunder batted, they made two.

When it came down to it, one team made an error off the last ball, and one team did not. Had it been the other way around, we'd have had a different result, but the data tells us that Morgan is a cool-headed chaser and Hilfenhaus is a second-choice death bowler. This time, the data was right.