I often find myself frustrated by the various fantasy cricket games that seem to score players on arbitrary, meaningless systems that don't really reflect the real game of cricket. They also often place restrictions on picking players, i.e. you need 5 batsmen, an all-rounder, a keeper and four bowlers. In real life (which in my opinion fantasy games should aim to replicate), a team isn't restricted in this way and players can change roles depending on what best suits the team (look at England's recent test teams). With this in mind, I designed a simple model based on the simple principles of test cricket.
To win a match of test cricket a team needs to make more runs and take the same amount or more wickets (unless there are some strange declarations) than the opposition. Bearing this in mind, I came up with the following formula to use as a scoring system:
(Wickets taken - Wickets lost) * (Runs scored - Runs conceded)
I thought this was a simple, yet accurate way of determining a player's performance in the context of a test match. The simple idea was that if a team took either less wickets than they lost, or made less runs than their bowlers conceded then their 'score' would be negative. Hence, the idea was to balance taking wickets and scoring runs, so a team of just bowlers would get a negative score, as would a team of just batsmen. Initially I was happy with this version of the model, although I realised that fielding was not taken into consideration and so the model would not pick a wicket-keeper unless they deserved to be picked as a batsman in their own right. A few tweaks were made to the model to encourage a keeper to be picked and to reward players who take catches. Catches were treated as 10% of a wicket, so a player who took 10 catches for the year was equivalent to a bowler taking 1 wicket. Also, a stumping was counted as a wicket for the bowler and the keeper to further enhance a keeper's value.
Next, all the test players who played in 2018 were put in a spreadsheet in Excel and the only constraints the model was told was there needed to be 11 players selected and no player could be chosen more than once. There was no need to tell the model it had to pick 4 bowlers and a keeper, I was confident the model would succeed in picking a balanced side because the scoring formula it was based on was designed to provide inherent balance to batsmen and bowlers. This was the team the model chose:
It did a great job at picking what appears to be a very strong team of players who performed well in 2018.
Unfortunately, the model did not select any players who played as a keeper throughout 2018, but both Mendis and De Villiers have kept in the past, and they took a lot of catches last year. A true weak point of the model being so simplistic is that it doesn't take into account any details on the performances of the players, as in it doesn't know whether all a player's runs were scored against a weak opposition of if all a player's wickets were against tail-enders. This is something, if I find more time, I would like to investigate to potentially weight certain performances based on the conditions.
When compared to the actual ICC test team of 2018, there are a few changes:
Karunaratne made the ICC team, but not my model's team, obviously the model didn't consider batting positions and hence didn't really pick a second opener, instead promoting Nicholls to open with Latham. Also, in the top six along with Karunaratne, Pant and Williamson are replaced from the ICC team for Mendis, Jadeja and De Villiers. Pant only averaged 38 for the year and so was really selected on the default that he was the best performed keeper of the year. Williamson is quite unlucky to not be selected but suffers from playing fewer tests than Mendis and Mendis' strong fielding record also helps aid his selection.
2018 certainly was the year of the fast bowler, with not much separating the elite fast bowlers. My model left out ICC selections Bumrah and Rabada despite their 48 wickets at 21 and 52 wickets at 20 respectively, but all of the fast bowlers it did select (Holder, Southee, Abbas, Cummins and Philander) had better bowling averages than these two. Also, Lyon was dropped from the ICC team by my model, with Jadeja the sole spinner and an extra seam bowler picked (with Holder and Jadeja both being all-rounders at 6 and 7). Lyon averaged 34 for the year and probably similarly to Pant, was picked because of the absence of other quality spinners throughout the year. Jadeja was in and out of the Indian team all season, which is probably why he was overlooked for the ICC team of the year. However, his record speaks for itself, with a batting average of 45.6 and a bowling average of 22.3, he is clearly deserving of selection.
It is nice to know that a simple model can be made to select a test team based on their statistics, obviously I'm not saying this should replace real selectors and it is certainly a very simplistic model, but it can provide some insight as to which players add the most value to a team. With some more time and detail added, this could potentially be used to create a proper cricket fantasy game based on what is actually needed to win a game of cricket.
Comments