Cribbage's Standardised Test Averages (UPDATED November 2018 - posts 753-755)

[weldone]

Batting[*] Adjusted longevity calculation for tailenders, now only giving players who bat 8-11 half the longevity credit for the match. This is to address issues we had with players averaging 20 at number 8 for 15 years being rated more highly than players averaging low 30s in the top order for 4 years.

Bowling[*] Adjusted longevity calculation for players who bowl less than 20% of their team's overs in the combined matches they play. For example, a player who bowled 10% of his team's overs in a career spanning ten years would get credit for 5 years worth of longevity on the bowling chart. A player who bowled 5% of the overs in this same career would get credit for 2.5 years worth of longevity.

In my ODI ratings I addressed the exact same problems in a slightly different way. Basically in our longevity calculation I replaced the number of matches by number of innings (for batsmen) or number of deliveries (for bowlers).

So, for a particular year, a batsman's longevity score is (number of innings played by him in that year / number of innings played by the batsman from same country who played the most number of innings in that year)

Similarly, for a particular year, a bowler's longevity score is (number of balls bowled by him in that year / number of balls bowled by the bowler from same country who bowled the most number of deliveries in that year)

 

[Prince EWS]

In my ODI ratings I addressed the exact same problems in a slightly different way. Basically in our longevity calculation I replaced the number of matches by number of innings (for batsmen) or number of deliveries (for bowlers).

So, for a particular year, a batsman's longevity score is (number of innings played by him in that year / number of innings played by the batsman from same country who played the most number of innings in that year)

Similarly, for a particular year, a bowler's longevity score is (number of balls bowled by him in that year / number of balls bowled by the bowler from same country who bowled the most number of deliveries in that year)

Yeah I'd probably do it this way for ODis too. I think you'd run into more problems with it in Tests though given the length of an innings isn't fixed, especially for bowlers (ie. it'd favour spinners and good bowlers in bad sides -- the same problem I had with my last method).

I'm happy with how I've addressed it now. I've always wanted to address it in this way but just didn't really have the data available for it with how my database worked.

 

[ankitj]

I think PEWS fudged the numbers for Goddard. Gets higher rank than Trueman and Holding.

 

[Cabinet96]

Wow Morkel at 30. Managed to get his final standardised average down to 23, and Anderson got his down to 22.

 

[Prince EWS]

Wow Morkel at 30. Managed to get his final standardised average down to 23, and Anderson got his down to 22.

It could actually change after he retires as the algorithm uses hindsight to determine the quality of the batting lineups he's bowled to recently, but yeah I can't see it moving from the 23-24 mark.

 

[Daemon]

Is Longevity for bowlers and batsmen the same? Seems odd that Cowie would get an 8.5 for 9 tests (3 in 1937 and the rest from 45-49) while Headley gets less than that

 

[Prince EWS]

Is Longevity for bowlers and batsmen the same? Seems odd that Cowie would get an 8.5 for 9 tests (3 in 1937 and the rest from 45-49) while Headley gets less than that

Other than the way it modifies them for bowling less than 20% of overs or batting in the tail, yeah it works the same way.

It's an interesting comparison though, and basically borne of a quirk in what happens when an entire team plays no Tests in a year.

Looking at Cowie:
First it calculates the percentage of games for NZ he played between the first and last calendar year he played - so that's from 1937 to 1949 - and he in fact played all of them (100%). On spec this would give him a longevity of 12, but as there were calendar years NZ played no Tests in it at all through that period, it goes through a bit of a different process, dealing with it somewhat year by year/
In 1937 New Zealand played 3 Tests, and Cowie played in all of them, so he gets 1 year's worth of longevity of that year - running total 1.
In 1938 New Zealand played no Tests. As Cowie played both before and after this calendar year he gets some credit for it, but only half his overall percentage (calculated at the start) - 0.5 * 100% which is 0.5 - running total 1.5
In 1939 same deal, another 0.5 - running total 2
In 1940 same again - running total 2.5
In 1941 same again - running total 3
In 1942 same again - running total 3.5
In 1943 same again - running total 4
In 1944 same again - running total 4.5
In 1945 same again - running total 5
In 1946 they played one Test, and he played in it, which is obviously 100% of the games. He gets a full point here - running total 6
In 1947 they played another Test, again he played, running total 7
In 1948 they didn't play, so again he gets half his overall percentage, which is half - running total 7.5
In 1949 they played 4 Tests and he played all of them, he gets another 1 full point - running total 8.5

Headley is a different case because in the calendar years he played, he only actually played 44.9% of West Indies of Tests. As such, in the years they didn't play any during his career, he was only getting half of that (0.2245) rather than half a full allotment, as it expects he would've played less than the full amount of Tests even if the West Indies were playing on. So his running total is as follows:

1930 - played 5 of 5, +1, running total 1
1931 - played 4 of 4, +1, running total 2
1932 - played 0 of 0, +0.22, running total 2.22
1933 - played 3 of 3, +1, running total 3.22
1934 - played 0 of 0, +0.22, running total 3.45
1935 - played 4 of 4, +1, running total 4.45
1936 - played 0 of 0, +0.22, running total 4.67
1937 - played 0 of 0, +0.22, running total 4.90
1938 - played 0 of 0, +0.22, running total 5.12
1939 - played 3 of 3, +1, running total 6.12
1940 - played 0 of 0, +0.22, running total 6.35
1941 - played 0 of 0, +0.22, running total 6.57
1942 - played 0 of 0, +0.22, running total 6.80
1943 - played 0 of 0, +0.22, running total 7.02
1944 - played 0 of 0, +0.22, running total 7.24
1945 - played 0 of 0, +0.22, running total 7.47
1946 - played 0 of 0, +0.22, running total 7.69
1947 - played 0 of 0, +0.22, running total 7.92
1948 - played 2 of 7, +(2/7), running total 8.20
1949 - played 0 of 2, +0, running total 8.20
1950 - played 0 of 4, +0, running total 8.20
1951 - played 0 of 4, +0, running total 8.20
1952 - played 0 of 3, +0, running total 8.20
1953 - played 0 of 5, +0, running total 8.20
1954 - played 1 of 5, +0.2, running total 8.40

Headley is probably the hardest and weirdest case in that coming back in 1954 might've actually cost him longevity (or gained him next to none - either way) in that then expected him to have played so much less in 40-47 when his team wasn't playing any games at all. It's just a really hard one to deal with without carving out a really special case for it, but ultimately the reason Cowie has greater longevity is that it's 'guessing' he would've played more in the years NZ weren't playing than it's 'guessing' Headley would've when WI weren't.

 

[weldone]

Ya I remember my initial thought was 0 credit is better than 0.5 credit for such years.

To me, asking the question 'How many tests George Headley may have played if there was no World War II?' is the same as asking 'How many tests WG Grace may have played if test cricket started 7 years earlier?' or 'How many tests Clive Rice may have played if there was no apartheid?'

The answer to all these questions is the same - probably many, but but they were unfortunate because World War II and Apartheid happened, and test cricket didn't start earlier.

 

[Prince EWS]

Ya I have said before why I believe 0 credit is better than 0.5 credit for such years.

To me, asking the question 'How many tests George Headley may have played if there was no World War II?' is the same as asking 'How many tests WG Grace may have played if test cricket started 7 years earlier?' or 'How many tests Clive Rice may have played if there was no apartheid?'

The answer to all these questions is the same - probably many, but but they were unfortunate because World War II and Apartheid happened, and test cricket didn't start earlier.

I think giving it 0 weighting just increases the reliance on the arbitrariness of a calendar year period though. If you treat 1935, 1936 and 1937 as separate periods then Headley has a longevity of 1 year in that timeframe, but if you treat that time as one period then he has a longevity of three years as he played all the games in it. Or you could go the other way and split it up into months, and then say that he didn't actually play any games in October '35 for example and as such doesn't deserve any longevity for it, only deserving longevity for the months in which he actually played. Taking these extremes all the way out, longevity just ends up either the amount of milliseconds between first and last game, or it just becomes the amount of games each person played. Either of these extremes defeat the purpose entirely.

I think I found a clumsy but workable middle ground with it, that works really well when teams play cricket every year, and *mostly* works even if they don't (other than in one or two cases like this one). I'll happily admit that the process doesn't give a great insight into how one should rate the longevity of Headley's career specifically though -- as I said a few months back his rating here basically lives or dies based on how I treat this problem, and as such he's a particularly subjective case.

 

[weldone]

I think giving it 0 weighting just increases the reliance on the arbitrariness of a calendar year period though. If you treat 1935, 1936 and 1937 as separate periods then Headley has a longevity of 1 year in that timeframe, but if you treat that time as one period then he has a longevity of three years as he played all the games in it. Or you could go the other way and split it up into months, and then say that he didn't actually play any games in October '35 for example and as such doesn't deserve any longevity for it, only deserving longevity for the months in which he actually played. Taking these extremes all the way out, longevity just ends up either the amount of milliseconds between first and last game, or it just becomes the amount of games each person played. Either of these extremes defeat the purpose entirely.

I think I found a clumsy but workable middle ground with it, that works really well when teams play cricket every year, and *mostly* works even if they don't (other than in one or two cases like this one). I'll happily admit that the process doesn't give a great insight into how one should rate the longevity of Headley's career specifically though -- as I said a few months back his rating here basically lives or dies based on how I treat this problem, and as such he's a particularly subjective case.

One year is a fairly long time-frame though. I mean come on - it made Steve Smith cry. Personally I'd give Headley 0 credit for 1932 and 1934.

Also, Cowie is a much bigger problem than Headley.

 

[Dan]

Hmmmm...I imagine Bob Simpson gets similarly nerfed with this. Is there a way to distinguish between those two retiring vs not being selected to not demolish their longevity ratings?

If there's ever two cases for manual adjustments...

 

[Prince EWS]

One year is a fairly long time-frame though. I mean come on - it made Steve Smith cry. Personally I'd give Headley 0 credit for 1932 and 1934.

Also, Cowie is a much bigger problem than Headley.

Cowie is a meme. "Less than ten Tests doesn't count" has always fixed this, eventually I just started to enjoy the meme though.

 

[Prince EWS]

Hmmmm...I imagine Bob Simpson gets similarly nerfed with this.

Nah he doesn't because it doesn't have to do any 'guessing' with him, as Australia played Test cricket every year during his career (including his first retirement phase). Playing 0 out of X just gets you 0 unless X=0.

It's a very Headley-specific problem.

 

[weldone]

Cowie is a meme. "Less than ten Tests doesn't count" has always fixed this, eventually I just started to enjoy the meme though.

Where does Headley end if you move from 0.5 to 0? (only if that's easy to calculate for you)

 

[Magrat Garlick]

Headley is probably the hardest and weirdest case in that coming back in 1954 might've actually cost him longevity (or gained him next to none - either way) in that then expected him to have played so much less in 40-47 when his team wasn't playing any games at all. It's just a really hard one to deal with without carving out a really special case for it, but ultimately the reason Cowie has greater longevity is that it's 'guessing' he would've played more in the years NZ weren't playing than it's 'guessing' Headley would've when WI weren't.

Fairly sure that method hits Headley ridiculously hard. If you stop the calculation at 1948 his overall percentage of Tests played becomes 75 %, which increases all the 13 empty years from 0.2245 to 0.375. So he would get 2.43 extra points for those, minus 0.20 for 1954.

Would it not be better to use the percentage for Tests in the two years either side of the gap?

 

[Prince EWS]

Nah he doesn't because it doesn't have to do any 'guessing' with him, as Australia played Test cricket every year during his career (including his first retirement phase). Playing 0 out of X just gets you 0 unless X=0.

It's a very Headley-specific problem.

Expanding on this, this is Simpson:
1957 - 2 out of 2, +1.00 - running total 1.00
1958 - 4 out of 5, +0.80 - running total 1.80
1959 - 0 out of 8, +0.00 - running total 1.80
1960 - 2 out of 5, +0.40 - running total 2.20
1961 - 8 out of 8, +1.00 - running total 3.20
1962 - 2 out of 2, +1.00 - running total 4.20
1963 - 4 out of 4, +1.00 - running total 5.20
1964 - 14 out of 14, +1.00 - running total 6.20
1965 - 6 out of 7, +0.86 - running total 7.06
1966 - 4 out of 5, +0.80 - running total 7.86
1967 - 5 out of 5, +1.00 - running total 8.86
1968 - 1 out of 9, +0.11 - running total 8.97
1969 - 0 out of 8, +0.00 - running total 8.97
1970 - 0 out of 6, +0.00 - running total 8.97
1971 - 0 out of 4, +0.00 - running total 8.97
1972 - 0 out of 7, +0.00 - running total 8.97
1973 - 0 out of 7, +0.00 - running total 8.97
1974 - 0 out of 8, +0.00 - running total 8.97
1975 - 0 out of 10, +0.00 - running total 8.97
1976 - 0 out of 4, +0.00 - running total 8.97
1977 - 3 out of 10, +0.30 - running total 9.27
1978 - 7 out of 9, +0.78 - running total 10.05

He plays until '68 with a longevity of 8.97 with no projections needed, then takes a few years off and comes back to add another year and a bit. Works perfectly fine for him.

It's a problem for Headley very specifically because he combined coming out retirement after his team had been playing Tests with earlier having missed a bunch of years due to his team not playing them. I don't even know if there is such another combination, and I'm not really sure how to deal with it exactly. Looking at the final result he seems to be rated okay IMO, but I think he's punished slightly for playing extra which really goes against the spirit of a longevity rating. Maybe I should code in a function to not count extra games towards longevity if they'll retroactively make it guess you would've played so many fewer earlier that you get less overall -- but I have a feeling it may apply literally just to Headley and no-one else.

 

[Prince EWS]

Would it not be better to use the percentage for Tests in the two years either side of the gap?

Perhaps. It doesn't really know that the gap is connected though, if that makes sense. It goes through it year by year, sees there are no Tests, moves on to the next year where there are no Tests etc. I could change that, but it'd probably be a couple of hours work for what I think could well just be one player, and it could have some interesting side-effects too. Lets say the year on the side of a gap is one Test in January or something and a player missed it, meaning it'll give him zero even though he played 7/7 the year the year prior -- that's a situation as bad as the Headley one, and it might pop up in a few different ways.

 

[Prince EWS]

Maybe I should code in a function to not count extra games towards longevity if they'll retroactively make it guess you would've played so many fewer earlier that you get less overall -- but I have a feeling it may apply literally just to Headley and no-one else.

If this was the case I guess I could manually calculate it and then hard-code it in to just change Headley's. Even finding out if it was the case or not could take a while though.

 

[srbhkshk]

Why not incorporate a factor that takes into account -
1. The number of years where one played at least one test.
2. The number of tests played.

I understand that #2 gives advantage to teams which play more tests - but given the bodily wear and tear associated with sports - it stands to reason that playing 10 tests in one year is somewhat worth more than playing 5.

Edit - People who missed out because of not playing tests because their team didn't play tests in those years should miss out, what value does a cricket player have if the team isn't playing cricket.

 

[Flem274*]

rather than all these amusing mental hoop jumps in a vain attempts to get rid of him, i think it's time everyone accepts Jack Cowie's rightful place as #1 test bowler of all time. Everyone who faced him raves about him, he's got the statzzzzz, looks fantastic on the youtube film someone dug up and it's not his problem he played for a crap team during a world war started by some european goons.

Jack Cowie forever, your favourites never. I believe in the debate to follow I'm meant to utilise the usual tactic of cherry picking and finding ways to call Marshall and McGrath secretly terrible so that's what I'll be doing. Also Lillee averages 70 in Pakistan. I haven't actually checked that, but it must be true.