DoG's Test Innings and Bowling Performances. Updates thread.

[capt_Luffy]

Can I ask what you mean by a 50/50 importance on home/away?

What if the match is played at home in 2002? Will the bowler’s home performances from 2000-2004 be measured? Or should I count all performances from 2000-2004? The former runs the risk of sample size lol issues and the latter has issues such as Jim Laker averaging under 15 from 1956-1960. He was a totally different prospect to face in England compared to anywhere else.

I honestly think I should just rate attacks on career home/away averages and be done with it. That irons out all peaks and troughs that players normally go through.

Yeah, that irons out the peaks and troughs and I think that's not really a good think. Pre injury Bishop, Botham and Waqar were completely different bowlers than post injury. I mean, think about it; was playing Malcolm Marshall in 1978 or 92 an equal challenge as playing him in 1987? I don't think so. I could see the problem in rating attacks on that split and Sample size problems; so I think in case of a bowlers home matches, you could just use their home average over say the last two and upcoming two years and/or last 20 and following 20 innings; etc. The problem really arises while rating away performances; in which case I can think of giving equal weightage to their home and away record over such a period (or might be slightly greater period). At the end of that, these are all just suggestions; I really like your work and way you rate attacks inherently doesn't really distorts the rankings.

 

[Days of Grace]

I have decided to go with 5 year rolling averages in combination with career averages. The latter smooths out the small sample size issues of the former.

The question is whether to use the previous five years or the two years before and after the year in which the match/series was played.

For example, for a match played in 1956, do I use batsmen/bowler records from 1952-1956 or from 1954-1958?

 

[Xix2565]

I think the latter sampling gives a better idea of the quality of the player in the moment.

 

[capt_Luffy]

I have decided to go with 5 year rolling averages in combination with career averages. The latter smooths out the small sample size issues of the former.

The question is whether to use the previous five years or the two years before and after the year in which the match/series was played.

For example, for a match played in 1956, do I use batsmen/bowler records from 1952-1956 or from 1954-1958?

I personally think that the later works better, as the former is only how good they were for the past few years, the latter takes in their form going on into consideration as well. Like, if a bowler had great 4 years and his fall began around the time the match was played, the latter accounts for it much better.

 

[Days of Grace]

Thanks for the advice. Five year periods surrounding the series in question have been implemented.

The second question is whether to further separate the averages into home and away/neutral?

One compromise could be basing a player’s rating on their career average together with their home or away/neutral form in the five year period. In essence we are answering what was the overall class of bowler the batsmen faced and what version of that bowler turned up.

 

[OverratedSanity]

Thanks for the advice. Five year periods surrounding the series in question have been implemented.

The second question is whether to further separate the averages into home and away/neutral?

One compromise could be basing a player’s rating on their career average together with their home or away/neutral form in the five year period. In essence we are answering what was the overall class of bowler the batsmen faced and what version of that bowler turned up.

I'd scrap the home/away/neutral thing entirely for this exercise tbh. Limiting it to five years and only considering home/away performances would limit the sample size and the problem I have with including home/away for bowlers in this rating is that they don't represent one "continuous" series of performances. There'll be series separated by multiple months/years with other games in between. Form is still form, both home/away performances should count imo.

 

[Days of Grace]

I'd scrap the home/away/neutral thing entirely for this exercise tbh. Limiting it to five years and only considering home/away performances would limit the sample size and the problem I have with including home/away for bowlers in this rating is that they don't represent one "continuous" series of performances. There'll be series separated by multiple months/years with other games in between. Form is still form, both home/away performances should count imo.

Good points. Also there is the issue of Pakistan batsmen and bowlers playing in the U.A.E. which would count as neutral performances. Also there is the problem of conditions varying greatly across different away/neutral venues.

I just thought I should differentiate between facing Jimmy Anderson at Melbourne in 2006 vs facing the same bowler at Old Trafford in 2018.

But there are a few too many sample size issues as you noted by splitting into home and away/neutral. In particular Michael Kasprowicz averaged over 80 away from 1999-2003, which really affects the overall rating of the Australia attack that Laxman faced in his 281.

 

[capt_Luffy]

Good points. Also there is the issue of Pakistan batsmen and bowlers playing in the U.A.E. which would count as neutral performances. Also there is the problem of conditions varying greatly across different away/neutral venues.

I just thought I should differentiate between facing Jimmy Anderson at Melbourne in 2006 vs facing the same bowler at Old Trafford in 2018.

But there are a few too many sample size issues as you noted by splitting into home and away/neutral. In particular Michael Kasprowicz averaged over 80 away from 1999-2003, which really affects the overall rating of the Australia attack that Laxman faced in his 281.

Is a middle ground not possible? As stated by many, away averages by country to country are a really small sample size and will definitely pose Kasprowicz like problems; as away averages may get down for a wide variety of factors and also the fact that away in Australia and away in India aren't the same, but similarly; facing Ashwin in India and facing him in SA have a monumental difference. Same for Jadeja, Murali, Ntini, Philander, Lillee, Laker, basically almost every bowler is much better at their home.

 

[Days of Grace]

Is a middle ground not possible? As stated by many, away averages by country to country are a really small sample size and will definitely pose Kasprowicz like problems; as away averages may get down for a wide variety of factors and also the fact that away in Australia and away in India aren't the same, but similarly; facing Ashwin in India and facing him in SA have a monumental difference. Same for Jadeja, Murali, Ntini, Philander, Lillee, Laker, basically almost every bowler is much better at their home.

I don’t do averages by country to country. At the moment the formula is:

1. Overall career average
2. Five year form average

Each are given equal weighting. The question is whether to split the five year form average into home and away/neutral. Any small sample size issues are ironed out somewhat with the career average.

 

[Coronis]

Is a middle ground not possible? As stated by many, away averages by country to country are a really small sample size and will definitely pose Kasprowicz like problems; as away averages may get down for a wide variety of factors and also the fact that away in Australia and away in India aren't the same, but similarly; facing Ashwin in India and facing him in SA have a monumental difference. Same for Jadeja, Murali, Ntini, Philander, Lillee, Laker, basically almost every bowler is much better at their home.

Not Warne or McGrath or Hadlee or Ambrose or Lyon or Gibbs or Garner or Benaud or Holding or Roberts.

Marshall confirmed HTB of the quartet @kyear2