cricrate: new cricket ratings website

[Red]

was talking to Lillian Thompson but his posts have been deleted.

Ah yeh, thought you were talking to me or viriya. Lillian Thompson does seem to have an aggro issue in this thread for no apparent reason.

 

[viriya]

cricrate

It's been 3 months since I first introduced the site.. Thought of giving an update on some changes since:

• Live match odds (shows up as a table at the top of the home page if a live game is ongoing): Calculates live match odds based on similar historical match results. Uses the live scores RSS feed from cricinfo - any relevant live game (Test, ODI, T20I, FT20, First-class, List A) should automatically display odds. Methodology here.
• Franchise T20 ratings: Separate format for the more competitive franchise t20 leagues (currently IPL, BBL and CPL). Similar to T20I in function, the idea is that a player's current rating after the IPL is transferred over when he plays in the CPL next, and so on.
• Better charting: Uses google charting for cleaner and more visually appealing charts, hover over the bars for additional info of each innings.
• Responsive styling to handle different devices: should render better in mobile and other screens now, a complete graphical overhaul is in the pipeline next which should resolve any remaining issues.
• Other minor adjustments (rating system is tweaked constantly aside from these changes):
  
  • Tests: Strike rate as minor factor, display strike rate where available
  • ODIs/T20s: Limit strike rate factor for short innings (6 and out - strike rate of 300 skews too much)
  • T20s: Lower wicket/run factor, remove wicket/ball factor, increase weight for economy rate (wicketless economical spells were undervalued)
  • Scorecards show target runs where applicable

Any feedback appreciated. If you think a certain rating is nonsensical, feel free to point it out as long as you think it's actual worth can be captured better using available data.

 

[BackFootPunch]

How is it possible that Hamilton Masakadza is rated the 23rd best Test batsman, given that he averages less than 24 against top 8 teams? If I remember correctly he scored a ton against the Windies at one point but all his other significant scores have come against Bangladesh. He' rated ahead of, among others, du Plessis, Pujara, Cook and Mominul Haque. That's mind boggling.

 

[viriya]

How is it possible that Hamilton Masakadza is rated the 23rd best Test batsman, given that he averages less than 24 against top 8 teams? If I remember correctly he scored a ton against the Windies at one point but all his other significant scores have come against Bangladesh. He' rated ahead of, among others, du Plessis, Pujara, Cook and Mominul Haque. That's mind boggling.

This is a current form index for established players.. Hamilton is in the form of his life.. he averaged 52+ with the bat 2014, even if it was vs Bangladesh, he was clearly Zim's best batsman. Pujara and Cook both had horrendous 2014s, and Faf is still establishing himself in Tests (the first 20 tests of a player are discounted in rating weight to ensure that a newbie with one good year doesn't shoot up to #1 in the list).

As you can see with Kohli's case, players can shoot up and down the ratings within a series - he had a horrendous England tour but has shot up 20 spots after 3 tests in Aus.. Hamilton's position can drop quickly if he doesn't continue with his good form.

 

[BackFootPunch]

Yeah but even if he's in the form of his life he's still a much, much worse batsman than someone like Faf, who's hardly a newbie anymore. I'd imagine Faf's played more games than Masakadza recently so surely his ability can be more consistently gauged?

I just don't see the point in showing the current form of someone if they're quite obviously not that good and their current form isn't that amazing. Averaging 52 against Bangladesh is hardly something to write home about.

 

[viriya]

The difference between their ratings is miniscule, and Faf will jump past as soon as the current test is done.

There is a tendency to completely discount performances vs minnow nations but it's somewhat unfair. The innings ratings are discounted because of the relatively low quality bowling attacks (although with shakib bowling a lot of the overs I would argue that the Ban attack isn't completely hopeless), but those runs still matter.

Another thing to note is that since Hamilton has barely any history of good performances, his current rating is more precarious than say Cook's for example.. Cook has had a whole year of poor performances and he is still rated just below Hamilton.

Hope that gives some explanation of the reasoning.

On whether I consider Faf's career to date to be better than Hamilton's - just check the Best Test Batting Careers list.. He's already made the top 100.

 

[NUFAN]

Kohli gained 182 points for scoring 115,141,19,1,169,54 against Australia.

Masakadza gained 151 points for scoring 158,61,81 and 38 against Bangladesh

 

[fredfertang]

How can it be that Graham Gooch's 154* at Headingley in 1991 doesn't figure in the top 100 innings?

 

[weldone]

The look of the website is good. But the methodology needs so much refining that I don't know where to start. Don't take it personally viriya, but I think the criticism you're getting is completely justified.

 

[weldone]

The all-time ranking methodologies need a complete overhaul. For a start, I think you should see Prince EWS' thread for all-time rankings.

 

[OverratedSanity]

How can it be that Graham Gooch's 154* at Headingley in 1991 doesn't figure in the top 100 innings?

The methodology places a very high weightage on the sheer quantity of runs scored. The smallest innings in the list is 153, and the majority are doubles and triples. The formula needs several more modifiers to make it a good one, imo. The base value of runs just overpowers the other criteria and makes it more a list of big, long innings.

I mean, any list of great knocks which puts ABdV's 278 on Dubai pancake in a draw at no.8 all time, and doesn't even feature Gooch's 154, Kim Hughes 100, VVS Laxman's 281, etc needs a lot more work.

 

[Spark]

I mean, any list of great knocks which puts ABdV's 278 on Dubai pancake in a draw at no.8 all time

I don't remember a single thing about that innings. I mean, I watched it, but it all seems to have vanished into the aether.

 

[Agent Nationaux]

What's the point when we already have DoG's ranking.

 

[Prince EWS]

Yeah but even if he's in the form of his life he's still a much, much worse batsman than someone like Faf, who's hardly a newbie anymore. I'd imagine Faf's played more games than Masakadza recently so surely his ability can be more consistently gauged?

I just don't see the point in showing the current form of someone if they're quite obviously not that good and their current form isn't that amazing. Averaging 52 against Bangladesh is hardly something to write home about.

As someone who dabbles in a lot of this stuff myself, the problem you get with Masakadza is not just that he played mostly against Bangladesh in 2014, but that Bangladesh played a lot of their cricket against Zimbabwe in 2014. In determining the quality of Masakadza's opposition, one must look at how their bowlers fared around the time he played them, and their stats won't be that bad for the very fact that they played Zimbabwe. For example, Bangladesh at home in 2014 didn't have markedly different overall bowling stats than South Africa at home in 2014:
Bowling records | Test matches | Cricinfo Statsguru | ESPN Cricinfo

So in fact, whenever Masakadza plays against anyone, that team's team stats for around the period he played them will be inflated by playing against his own poor side, so any such system will tend to slightly (or not so slightly) over-rate players from weak teams. This problem caused me to rejig the entire way I determined the quality of opposition batting lineups and bowling attacks when standardising performances, and it took me longer than anything else in the entire system. so I can see why it's a problem for viriya.

 

[viriya]

Kohli gained 182 points for scoring 115,141,19,1,169,54 against Australia.

Masakadza gained 151 points for scoring 158,61,81 and 38 against Bangladesh

Although it seems that Hamilton gained a similar number of points for performances vs Bangladesh, that's actually not the case.

Kohli went from 445 to 627.
Hamilton went from 385 to 511.

Since Hamilton was rated lower to start with, he can gain points easier than Kohli, who has to perform at a higher level to gain the same number of points. Even then, Kohli gained more points because as we know, his Australia performances have been up there.

 

[viriya]

How can it be that Graham Gooch's 154* at Headingley in 1991 doesn't figure in the top 100 innings?

The methodology places a very high weightage on the sheer quantity of runs scored. The smallest innings in the list is 153, and the majority are doubles and triples. The formula needs several more modifiers to make it a good one, imo. The base value of runs just overpowers the other criteria and makes it more a list of big, long innings.

I mean, any list of great knocks which puts ABdV's 278 on Dubai pancake in a draw at no.8 all time, and doesn't even feature Gooch's 154, Kim Hughes 100, VVS Laxman's 281, etc needs a lot more work.

I don't remember a single thing about that innings. I mean, I watched it, but it all seems to have vanished into the aether.

It just misses out with a rating of 2657.

I hear the criticism here and I've been attempting to solve this issue myself. Just because I have certain innings up there doesn't mean I'm tied to them btw - I just want to get it right.

I agree that the runs weight seems too high when you look at the greatest innings compilation, but the reason that is the case is because if I lower it any more significantly, players like Jack Hobbs will drop lower below #24 in the Career rankings. This is because even though he made a lot of runs, the quality of his opposition wasn't that high.

I myself believe that Lara's 153 and Gooch's 154 should be rated higher, so I think I will tweak the base runs factor slightly for that reason.

 

[viriya]

The look of the website is good. But the methodology needs so much refining that I don't know where to start. Don't take it personally viriya, but I think the criticism you're getting is completely justified.

Thanks for the comments. I really appreciate any feedback - what I hate is none. Feel free to skewer the website as long as you have workable suggestions

 

[viriya]

What's the point when we already have DoG's ranking.

DoG's ratings works very well when rating ATG innings, but I'm willing to bet that it would be impossible to implement as a standard formula for all innings ever played and then for rating careers as mine does. I'm fairly certain that you would get batsmen who made useful runs when the team was in strife rated a lot higher in the ATG rankings (say lower order bat that averages 35 over a 50 average batsman)..

His ratings formula is impressive though - I stole a couple of ideas for mine (support and closeness factors).

 

[harsh.ag]

It's a bit weird about AbdV. Great batsman, great to watch, but not as memorable as many other greats, or even Amla for that matter.

 

[viriya]

So in fact, whenever Masakadza plays against anyone, that team's team stats for around the period he played them will be inflated by playing against his own poor side, so any such system will tend to slightly (or not so slightly) over-rate players from weak teams. This problem caused me to rejig the entire way I determined the quality of opposition batting lineups and bowling attacks when standardising performances, and it took me longer than anything else in the entire system. so I can see why it's a problem for viriya.

Good point, but that effect should be relatively minor. The Bangladesh bowlers (aside from Shakib) would be rated pretty low to start with that 1-2 tests of good performances (against inferior opposition, so those performances would be discounted as well) won't make much of a difference going into a third test. Interesting to think about though.

The way the current form of a player is determined is by an exponential smoothing method where more recent performances are given more weight than older performances. Even still, a player's debut test is still part of the calculation, just very minor compared to his most recent performance. This way, great players don't drop like stones when they have a patch of bad form, and inferior players don't maintain their ratings jumps easily.