Cricket probabilities

[sohum]

Whilst in the path of designing a simulation engine, I have run into the problem of probabilities. Not using them, but estimating them. Since there are a lot of experienced cricket viewers here, I'd like to poll the waters as to what you think applicable probabilities would be in an OD game, disregarding factors such as pitch, ground, field setting, etc.

Currently I have it set up like this:

Equally Matched Batsman/Bowler
0 runs: 62%
1-2 runs: 24%
3 runs: 5%
4 runs: 5.5%
6 runs: 1.5%
Wicket: 2%

Bowler Much Better
0 runs: 75%
1-2 runs: 4%
3 runs: 0.5%
4 runs: 0.4%
6 runs: 0.1%
Wicket: 20%

Batsman Much Better
0 runs: 18%
1-2 runs: 52%
3 runs: 7%
4 runs: 14%
6 runs: 8.5%
Wicket: 0.5%

There are a few other categories, but I'll adjust those based on your recommendations. Basically, in any given ball, what is the probability of a batsman getting X runs/wicket above?

It would be great to have statistics such as dot balls/singles/doubles/triples/fours/sixes/wicket deliveries in an average OD game, so if anyone can point me to anything liek that, it would be great too. Thanks!

 

[nightprowler10]

62% seems a bit high for equally matched batman/bowler. Something closer to 50% would seem more realistic.

 

[sohum]

Hmm, yeah, I played around a bit with it. When I set dot balls up to around 64%, the scores were realistic. Otherwise, there are far too many singles/boundaries being scored, which, ironically, boost the score a lot.

 

[Iron-Haggis]

Your equally matched batsmen/bowler look good, but the others seem a bit off in my opinion.

With your bowler much better than batsmen you're saying that an average over would be (. . . 1 W .), so 1 wicket for 1 run against a tail ender in a one dayer, when they'd be going for a slog.

With the batsmen much better than bowler you're saying that an average over would be (. 2 1 4 1 6) 14 runs with no wicket, maybe in the last 10 overs.

Both those seem a little off.

There are quite a lot of variables you need to consider to make a realistic simulation cricket game.

 

[sohum]

Yeah, I've taken into account a few other factors, which ultimately provide me a difference between bat and ball. These factors include runs required, run rates, overs left, wickets left, etc.

I tweaked the probabilities above by looking at the breakdown in a few cricket games which I got by painstakingly adding stats from the cricinfo's ball-by-ball breakdown (actually not that hard). The simulation engine is intended to be simple because it will be part of a bigger game, the focus of which is not hte actual gameplay. My simulation engine currently gives a big range of scores from about 180-280 on average. There have been a few 300+ scores and a few low scores. A realistic distribution, I feel. Probably not that close compared to a commercial alternative, though.

 

[Iron-Haggis]

How many wickets do you usually have falling? 180 to 280 seems pretty good for simulation. Get many people scoring centuries? Or many 5 wicket hauls?

 

[sohum]

I am still developing the engine, as I have only tested it extensively with two teams (India and West Indies). But here are a summary of scores for 10 consecutive games. As you can see, it's still pretty random, but suits the purposes.

1. WIN 264/6, 50 (Sarwan 101, Harbhajan 4/47) beat IND 202/10, 46.3 (Tendulkar 50, Bradshaw 3/46)
2. WIN 250/5, 50 (Gayle 65) beat IND 190/10, 39.2 (Gayle 4/45)
3. WIN 175/10, 41.2 (Harbhajan 3/27) lost IND 176/5, 31.1 (Raina 88*)
4. IND 255/10, 43.3 (Samuels 5/44) beat WIN 225/10, 45.3 (Sehwag 3/32)
5. IND 282/4, 50 (Yuvraj 108*) beat WIN 238/8, 50 (Chanderpaul 104*)
6. IND 321/5, 50 (Yuvraj 93*) beat WIN 115/10, 31.5 (Harbhajan 4/29)
7. WIN 235/8, 50 (Gayle 94) lost IND 238/3, 44.3 (Sehwag 131*)
8. WIN 241/7, 50 (Gayle 114) lost IND 242/0, 43.5 (Sehwag 170*)
9. IND 245/5, 50 (Dravid 65*) beat WIN 217/7, 50 (Harbhajan 4/18)
10. IND 203/10, 39.4 (Yuvraj 59) beat WIN 140/10, 28.1 (Pathan 4/51)

I've bolded the ones that don't seem real. Only two that seem to be kinda iffy, the other ones could be actual ODI results. The various factors I'm using are preassigned ratings, form, ability to play the type of bowler that's bowling, difference with required run rate (while chasing), over number (more attacking towards beginning and end). There may be a few more.

 

[nightprowler10]

I'd say number 6 is a likely result. Many teams fail miserably when chasing after a big total.

 

[sohum]

I guess... I'll try it out putting in some other teams with different ratings.

 

[Iron-Haggis]

Maybe a good idea will be to do Bangladesh and maybe Australia or South Africa, be a good way to judge how well your ratings work. Because there is a huge difference between the teams, but India and West Indies are very evenly matched in ODIs.

 

[sohum]

True, but I haven't added those players to the database yet. I added Australia and they beat India pretty consistently. Again, it's also doing stuff on form. For example, Australia win 4-5 in a row, but after a few close matches, the Indians came back and won a couple or more.

 

[w3lshrugby]

Nice work sohum, are you going to be releasing this engine?

 

[sohum]

Nice work sohum, are you going to be releasing this engine?

EDIT - The 600 is up :onpc

Probably not. It is more an in-house project.

 

[slade]

what about no 8, sehwag 170 NO?! 241 is a pretty good score and for india to chase it down without losing any wickets seems a bit unrealistic with 6 overs to go.

 

[rickyp]

gotta agree with slade there, no.8 is very unrealistic. Overall, nice work sohum, really top stuff