DoG Top 100 Test Batsmen : How much importance should be given to strike rate?

[RossTaylorsBox]

You can calculate error in a supervised learning model because you have a ground truth. But here there's no objective error function, it's literally just what you personally think is good or not.

 

[RossTaylorsBox]

The other approach is to create an "objective" ranking list and then people train models to try and predict it from statistical features.

 

[honestbharani]

You can calculate error in a supervised learning model because you have a ground truth. But here there's no objective error function, it's literally just what you personally think is good or not.

Yeah, unless the model actually makes an objective error somewhere, its just a case "its not matching my opinion, so its wrong"

 

[honestbharani]

The other approach is to create an "objective" ranking list and then people train models to try and predict it from statistical features.

But even then, the interpretation would depend on the original bend the programmer gives it first time, right?

 

[RossTaylorsBox]

Yeah that's why I quoted objective, you would have to have a consensus on the ground truth list. No impossible, but getting a consensus on this website lmao forget about it

 

[capt_Luffy]

The other approach is to create an "objective" ranking list and then people train models to try and predict it from statistical features.

I don't.... disagree. But in a model, any weightage you put in a parameter, is rather arbitrary. I remember a statistical analysis here, that had Barry Richards ahead of Don. Can say that's how that model worked, but can be said it was **** for determining the best Test batsman.

 

[honestbharani]

I think the underlying issue with any stats based exercise for cricket rating is the fact that so much of cricket stats is contextual which is never captured absolutely in the numbers.

 

[RossTaylorsBox]

I think something interesting would be to run like a PCA and then cluster batsman.

 

[ataraxia]

There are no hard lines there imo.

There are also no hard lines between sea-level air and outer space.

 

[RossTaylorsBox]

I don't.... disagree. But in a model, any weightage you put in a parameter, is rather arbitrary.

A supervised model would work out the parameters for you.

I remember a statistical analysis here, that had Barry Richards ahead of Don. Can say that's how that model worked, but can be said it was **** for determining the best Test batsman.

I actually disagree with this and believe it's completely fine for it to happen lol

 

[capt_Luffy]

I actually disagree with this and believe it's completely fine for it to happen lol

Idk tbh. All metrics and models aren't equal, even when there are no proper "objective" answers

 

[Two short legs]

While I'm happy in-principle for the inclusion of strike-rate, I worry that in practice it has the potential to skew results if given too high a weighting.

Hammond was happy to trundle along at 35 runs/100 balls during 1928/29. Timeless matches, high over-rates and flat pitches made it the perfect risk-reward outcome. Should he be penalised for having batted to the prevailing conditions ?

 

[RossTaylorsBox]

Idk tbh. All metrics and models aren't equal, even when there are no proper "objective" answers

I guess I should clarify that if you make a model with no objective way to measure it, you can't claim it as being authoritative, good, or even compare it to others (this is the reason Narayan sucks btw). Once you start doing this anyone can call it dogshit and pick it apart.

 

[ashley bach]

On one hand SR kind of seems to matter and on the other it makes jack all difference. It can only really be properly interpreted given the context
of the match and if scoring quickly was needed to benefit the team. Good luck trying to work out players who over their career scored at a faster
rate when it was required.
I would give rating points to players who have a better SR but only a very minimal amount. A very very minimal amount.

 

[Jane Austen]

Please vote on the following for the next 24 hours:

A: Keep the same weighting (Average x4, Runs per innings x 2, Strike-rate x1)
B: Less emphasis on strike-rate (Average x6, Runs per innings x 3, Strike-rate x1)
C: No strike-rate (Average x2, Runs per innings x1)

A fervent,even desperate plea to you Days of Grace.
Even if popular demand requests the diminution or removal of strike-rates from the evaluation of batters,please do not give way to this request.
Embedded in the fascination of Test,and first-class cricket,is the coming together,the equation,of runs,wickets AND time.Take away that vital third element of time is akin to taking the saxophone out of BAKER STREET by Jerry Rafferty.It's the time element which can,and often has,given Test cricket the neccesary vigour to what would be no more interesting than watching two tortoises race a marathon.
That vital ingredient of time,of itself,consequently lays emphasis on the amount of runs scored in a session,a day,and throughout the match and will ultimately have a bearing on it's result.
No one is decrying the greatness and the need for the batters of ballast,the grafters who lay the foundations of the innings.Their great skills of concentration,playing within the limit of their ability and being risk averse are rewarded and statistically recognised by the standard averages.
Accordingly those who are chosen because their greater range of stroke will give them licence to increase the scoring rate and,by definition,increase the possible chance of winning rather than just drawing the match,should have their different skill-set given credit by virtue of strike-rate being part of any assessment of batting comparisons.
Those who seem to be horrified by any credit being given to a faster rate of scoring have,so far,not given any logical objection as to why this particular part of a batters skill-set should be ignored.
To repeat---The faster the scoring rate the more chance of WINNING a match so,if possible,a stroke-player or two of appropriate quality should be considered.
Of course there must be a balanced batting line-up to include quality accumulators who will do their job of binding the innings together AND guiding their team to a draw if need be.
However the batters and their different skills aren't mutually exclusive.They bat together as a unit and in accordance with trying to win and,if this isn't achievable,seeking the draw.
But the basic premise is that the faster the scoring rate,the better chance of winning,thus strike-rate has to be given due credit in the assessment of a batter.This has been recognised by historians hence why faster scoring batters with lower averages have been considered greater than their contempories who have a higher average (to the shock-horror of some on this forum!)
Some on this forum need to look up from their spreadsheets and watch a game of cricket as well as reading the books of the great cricket historians.
So Days of Grace,please don't be swayed by an illogical cry to ignore strike-rates-------stay true to your own instincts and beliefs.....PLEASE.

 

[Randomfan]

SR is important as far as player has ability to change gears. Some were incapaple and some had capacity. Batsmen, who could change gears based on team's need were simply superior than some who did not have skills to do it if other things were equal. That part is obvious to most people. Other part is less obvous but plays a part as well. That part is dominating batsmen can derail the opposition attack.

But I am not sure how much SR was important lets say before 2000. It probably had less impact on results. Some one like Sehwag has won games by simply batting at very high SR on totally flat wickets to allow more time to bowles to pick 20 wickets.

After all said and done, list is stats based. It's fine to say that here is ranking based on this and this. It can differ from consensus or personal liking. Great effort by DOG here. Thanks for doing it.

 

[Johan]

SR is important as far as player has ability to change gears. Some were incapaple and some had capacity. Batsmen, who could change gears based on team's need were simply superior than some who did not have skills to do it if other things were equal. That part is obvious to most people. Other part is less obvous but plays a part as well. That part is dominating batsmen can derail the opposition attack.

Batsmen who play faster aren't inherently more skilled, batsman who have more range are more skilled. I don't see any evidence someone like Brook can play like Boycott, therefore isn't it fair to say he is equally one dimensional?

 

[Randomfan]

Batsmen who play faster aren't inherently more skilled, batsman who have more range are more skilled. I don't see any evidence someone like Brook can play like Boycott, therefore isn't it fair to say he is equally one dimensional?

If batsman has no capacity to change gears they are inferior than players having capacity to change gears.
Yes, if you can't play slow any time despite team needing it then you are one dimensional.
One dimnesional player from one category can be better than one dimensional player from another category.

 

[Johan]

If batsman has no capacity to change gears they are inferior than players having capacity to change gears.
Yes, if you can't play slow any time despite team needing it then you are one dimensional.
One dimnesional player from one category can be better than one dimensional player from another category.

No, all one dimensional players are equal as far as scoring tempos go and don't deserve to be rated higher or lower, no matter if their scoring rate is 35 or 75. Now, if a player can play both 35 and 75 depending on context, then sure, he deserves credit for versatility and an ability to adapt, otherwise no.

 

[Randomfan]

No, all one dimensional players are equal as far as scoring tempos go and don't deserve to be rated higher or lower, no matter if their scoring rate is 35 or 75. Now, if a player can play both 35 and 75 depending on context, then sure, he deserves credit for versatility and an ability to adapt, otherwise no.

I disagree on that. I will see that on case by case basis and not as rule.

As rule, ability to adapt is better than always slow scoring or always high scoring without any ability to change gears. But ability to adapt is harder to capture in career aggregate stats. For models, it will be hard to do.