All4Golf

Most players have no argument with the handicap system and would like it to work. The biggest negative is how incomplete and underreported are. Not every score is reported (both home course and away play). When 10-20 scores are in, the handicap system numbers are not going to change too much so there is no incentive to keep it up to date. Golf courses with 'competitive leagues' handicap (separate from USGA) are often lower than the player's course handicap. Todd All4Golf

I look forward to seeing how things work out in regards to the 40-30-20-10 rule. Please purge the content of my threads here. Todd All4Golf

We can go one of three routes for a successful proof of concept: A) It really works as is and have data to significantly back it up for all components of the model. B) The model is improved and better than original concept. C) The model does not work for 2006 tour data with reasons why. It could be due to reasons of statistical insignificance, ranked nature of the data, or how the data was analyzed. My take is option B - it can get better. It's ok with me if you want the concept to be left alone as is. Regards, Todd All4Golf

Being objective here, I've yet to see a well reasoned reply to the optimized weighting of ranked stats. It has a much better predictive relationship of the player's actual finish in the 2005 Tour Championships by statistically eliminating Driving Distance and Driving Accuracy as variables to the model. There is no denying how it presents a legitimate challenge to the overall validity of the rule. Common sense says that driving distance and accuracy are important but if the data doesn't validate it - unwanted noise is being added . Just be prepared to walk the line of real data validation. If factors are deemed statistically insignificant, recognize it and potential reasons why. Don't just throw 20% DD 10% DA or 30% weighting because it makes common sense. Perhaps DD or DA info or comparable info needs to be introduced to the model in a different manner. Regards, Todd All4Golf

It is hard to believe that the Editor in Chief of this website would say something like this. NM Golf deserves an apology because all he was doing was sharing an idea and his practical experience in tracking stats. Stats gathered are most useful when it has a purpose and all NM Golf was doing was trying to make his data more meaningful so he can improve. As long as he is consistent in his approach and interpretation, he'll do fine. There are plenty of golf stats that are rather negative about a person's game until he or she can get to single digit handicaps or better. Some say golf is a game of confidence because the game is frustrating enough as it is. iacas started this thread with a simple message of "I keep stats all the time. In fact, I'm still helping to write some software that keeps stats the way I'd like." NM Golf has his own way and incorporates some judgement in a few stats to make the data more meaningful and practical. The objective is the same - to become a better golfer. Regards, Todd All4Golf

I do and find it contributes to a positive outlook when you start a hole. Regards, Todd All4Golf

I appreciate Dave's attempt to offer some new insight and correlate available stats to a player's actual finish. It is a refreshing presentation and am curious to see how it works out with 2006 tournament data. Look forward to new 40-30-20-10 developments as they occur. Regards, Todd All4Golf

It would mean something only if we can assume the DD environment of 2 holes applies to the 12-14 other holes, which is impractical when navigating doglegs, bunkers, trees, and other obstacles. How do we know if the golfer uses a 3w instead of a driver on a DD hole? Trying to get Driver Distance to mean something appeals to every golfer's common sense. The caveat is in how it was collected to support the overall data analysis objective: Is there a way to use golfer's performance rankings to predict a player's actual ranking of a golf event or season? Didn't you just say that a longer player has shorter approach shots ? This is exactly why I said that closer to green translates into higher GIR. The GIR success decreases the further we are from the hole. Even ShotLink data backs it up with increasing average distance from the pin . Many folks will agree the graph looks good. Most will not understand the nature of the data to realize it reveals nothing unusual when rank ordering and summing two correlated variables to a low score - GIR and Putting (Total Putts, Putting Avg). Data simulations will prove it. It will appear to have a linear middle, with sharp deviations in the beginning and end. A more practical approach reflecting the actual golf process can make the limited golf data collected more valuable. I mentioned earlier, a clue to heart of the problem goes back to GIR because it covers shots over a wide distance range on the course. We know nothing about how precise the shots are to the green and how the player is succeeding in scoring opportunities. These little differences can make a huge difference in making more sense of the data and what it is that drives players to be successful on the tour or local course. Regards, Todd All4Golf

Let's consider the tour championship dataset. If we were to optimize the weighting among the 5 factors (GIR, Putts(PA, TP), DD, DA) relative to the player's finish with a rule that no one factor can be smaller than 0 and larger than 1. The optimal formula for the Tour Championship data comes out with a huge drop in variation: Estimate of player's finish = .44*GIR + .17*PA + .43*TP + 0*DD + 0*DA or 44-60-0-0 or simply 44% of Player's GIR ranking and 60% from putting rankings (or 17%*Putting Avg ranking and 43%*Total Putt Ranking). In other words, stats collected for Driving Distance and Driving Accuracy have no value whatsoever in the optimal model to match the actual player rankings. This kind of makes sense because distance remaining to hole or approach shot distance would be more meaningful than driving distance (ex: Closer = higher GIR, further away = lower GIR). Also, information regarding how players keep it in play off the tee is not readily available either. If data analysis says certain data is 'noise', mining it for useful information won't yield very much. A few pieces of the puzzle are missing for the overall picture to make more sense. Todd All4Golf

20% and 10% seem about right to me. It is a fact that driving distance is measured only on 2 holes per round: Driving Distance data as a fraction of Driving Accuracy's 14-16 holes is hardly representative of the overall golf round. Does it make sense to to have Driving Distance to be 100% more weight than Driving Accuracy when Driving Accuracy has 7x or 8x more data? Hence, some people say the driving distance is greatly overrated in a round while getting more than it's share of media hype. I agree both are important - Driving Distance shortens the course and makes it easier. Driving accuracy helps too. Is it also important to know whether ball is in play to the green or not? A player can be in the wrong side of the fairway on a dogleg and not have a direct shot to the green, or be sitting behind a tree in the middle of the fairway at Pebble Beach #18. A closer look at the SONY Open had countless of players in not in play situations. For these reasons, it would be easier for any golfer to apply the following reasoning to their round - 40% GIR, 30% Putting, and 30% off the tee in play performance than get too technical with DD and DA data. Or simply focus on two things - 70% how the hole is finished and 30% keep it in play. Todd All4Golf

Dave, We're in agreement here regarding GIR and putting to pinpoint issues with a player's game. The first item was not proven. I'll rephrase it differently: If golfer A has a 75% GIR and golfer B has a 70% GIR, can we conclude golfer A is statistically different than golfer B for a single round of golf (with a 5% margin of error)? For a single round of golf, there is a 33.84% chance the 70% GIR golfer will perform better than than the 75% GIR golfer. Therefore, we cannot conclude golfers's GIR are statistically different. How about 4 rounds of golf? The odds drop a little to 20.23% that the 70% GIR golfer's 4 round GIR average will beat the 75% GIR golfer's 4 round GIR average. Therefore, we cannot conclude golfers's GIR are statistically different. How about 16 rounds of golf? In this case, the 75% GIR golfer's 16 round GIR average performance will beat the other golfer 95.2% of the time. The margin of error has dropped to 4.78% and can statistically conclude that there is a difference. (footnote: GIR stat is a proportion measure. See this link for details on making inferences regarding proportions http://www.ed.uiuc.edu/courses/epsy480/notes/l2122.htm ) So what does this mean for a four round tournament and the 40-30-20-10 rule? The difference of 5% between two golfer's GIR doesn't mean much statistically among a group of highly talented golf professionals and we are unlikely to see 16 round tournaments in our lifetime. Perhaps using 4 years of golf data at the same course can make it more clear what it "is" that drives results of top players. Further study would be beneficial to see how it holds up for mulitple tournaments. It should be no surprise to see a course discrepancy (ex: Masters vs British Open) where the weighting could swing from 40-30-20-10 to 30-30-20-20. If such discrepancies occur, then a season average of the weights from every tournament be used to form the basis of a general guideline. The approach of ranking them and applying a 40-30-20-10 weighting scheme is a creative one to form a clearer alignment of player performance with the actual results. Keep working at it through out the 2006 season! I mentioned earlier the stats associated with GIR, Driving distance & accuracy info are not used and get better results. What we have in common is the basic fundamentals of "in play" and "finishing the hole" stats data. For example, relating to the 40-30-20-10 rule, the 40-30 weighting component is another way of saying of how they finish the hole, and the 20-10 component is how they keep it in play. what is different is data used and analysis. Ricco's rule may be of interest. He tied GIR with the player's score as a function of 95 - ( 2 x number of Greens hit in Regulation)? This result is expected to be within 1-2 shots of the final score when shooting below 95. Perhaps using this rule can filter out some some unusual data for a better 40-30-20-10 relationship. I don't use this measure but not sure if you've heard of it and thought it may be of value as part of the GIR emphasis. Where do you go to for most of your stats? That'd help out to make sure we're working with the same source of data. Having whole slate of PGA stats would be nice but not necessary. Too many things are being tracked, in my opinion, because a few things matter the most to achieve real and measureable improvement in a golf game. Regards, Todd All4Golf

Thanks for inputing a few comments. I'm sure Dave will chime in soon. The use of rankings DO take out any bias by player for a given course. I have no problem with that. I'm curious how does the 40-30-20-10 rule hold up for different courses and average score significantly below or above par ? Will it reveal anything about golfer playing styles that make them successful at a tough course ? Finally, can this rule tell a player how good their game needs to be to win golf tournaments or at least be in the top 5 ? (Ex: GIR needs to be at least 75%, etc...) "If a pro ranks low in GIR or Putting, then he can look at why he is bad. If he's ranked high in putting but he hits a lot of GIR, is he a bad putter or does he leave himself long first putts?" Good points. At some point every pro realizes that there is only so much he can do with analysis of GIR and putting info. It is easy to overdo it and overlook the positives that happened. Todd All4Golf

It is a rule that "sounds" nice and well reasoned with supporting graphs. Golfers can grasp it quickly and appreciate your efforts here. The most problematic stat for most golfers is GIR because it varies the most. If you can answer how big of a difference GIR has to be between one player and another to be deemed statistically significant, the tip of the iceberg of statistical issues have been discovered. Probing further into sources of variation with GIR will reveal how it covers the broadest distance range of shot making (ex: from a 300 yard fairway shot to hit green in two, to around green chip shots from driving a par 4). There are a lot of variables in play affecting the success & failure of GIR measure that lends it to be one of the most frustrating stats in golf. Golf course set up (ex: Baltusrol, or classic British open style courses) can easily throw the 40-30-20-10 numbers out of whack. Another confounding factor is the style of play of pro golfers. Tiger Woods plays courses quite differently than Corey Pavin. Golfers like Kenny Perry and Colin Montgomerie are play the game in pure and simple scoring style. To discover "what it is that drives the results of tour players" needs a different approach with data that is not easily available. I have a possible solution unlocking these secrets and don't pay any attention to driving distance or accuracy or GIR. The results? It is more clear what it takes to win a tournament. How can this be possible when the 40-20-10 is thrown out the window? The key is in minding the data, not mining the data. Regards, Todd All4Golf