Reply to thread

Message: [QUOTE="FireballPhil, post: 4619261, member: 553"] Okay I was gonna be done and let you just be a fool but for some reason i just can't help myself. I guess it is the part of me that always wants to see people improve and get better. Being able to track statistics and actually understand how statistics work in real life applications is two totally different things. You may feel you have a good handle on this but your statement "Even if there is some initial rating on the "power element" of the NET formula (the BPI), like KP the initial ratings in BPI simply roll off as the data becomes connected after about half a dozen games are played." shows you have zero idea of how these actually play out. Before I go any further let me state that I work in the meat packing industry, which is a business that uses statistics every single day. There is a lot of money made and lost when people follow or don't follow where the numbers lead you and react accordingly. Those that don't pay attention to them don't retain their jobs for very long. Statistics are very easy to chart, and data can be collected in multiple ways, it can also be manipulated in various ways. The bottom line is that most of these need a solid baseline. In the meat industry every day is like the start of a new season. How you start each day, in our company even how you start each shift is critical to the overall success (money making ability) of the plant. The real truth is the further into the day you wait to make any changes the smaller the overall impact becomes. This is exactly opposite of your view that the data roll off as games are played. Data doesn't just roll off, it is incorporated and combined give you a bigger picture. But that bigger picture doesn't always give you the truthful picture. You get this by other reports giving you actual total bodies of work. Inevitably this leads to a category that we call "unaccounted for" because the two numbers never show the exact same picture. I am getting a bit offtrack so let's just use giveaway as an example since it runs fairly close to the NET. When you go to the store and buy a prepackaged package of luncheon meat it may say on the package it is 1 pound. If you took the meat out of the package and weighed it and it came to 1.1 pounds then congratulations, you just hit the jackpot and my company just lost money. Potentially a lot of money. If we start the day giving away .1 pound every single package over the entire course of the day assuming no corrections were made we end up giving away hundreds of pounds of meat. To help avoid this we use statistics derived from various sources throughout the run. If the first statistical check shows we are running heavy, a change to the process must occur. If you do and the next check is good, the average of those two checks drop the amount of lost product in half. If you wait until the third check before you make the change you only drop the amount of lost product by a third. In other words the longer you wait in the process to make a change the less overall change you statistically show. Now change this to a bball season. If you start 9-0 then you have a pretty good average to start the year right? If you then start to lose a couple games then your average begins to come down but very slowly. The question really is how good are those 9 teams that each team played and how do you compare them? Whoever programs the NET has a predispositioned bias because someone has to start a ranking somewhere. Quadrants are assigned, perhaps incorrectly. As games, the season progress there are indeed corrections made and it does start to show more accurate data. The longer season, the more games played now as compared to years ago actually do help metrics such as the NET and make it more accurate, but it takes a very long time. This is no different if I am not getting the changes on the giveaway needed at my plant I can take more readings and basically force the system to show a downward trend. It is all simple math. The problem becomes when I have accounting run a simple pounds in versus pounds out report there can be a very big difference between the two. That "unaccounted" factor now comes into play. I get that this is a simplified, perhaps even "dummied down" example because it is a bit more complicated than this because of other outside factors but this is a good example of how metrics are great but must be used with a grain of salt, knowing that there are built in biases and never truly show the true full picture. The NET is a great tool, and by all means continue doing all your research and tracking. But please don't show your ignorance or assume you are the only one who understands your metrics or how they work. There are a lot of people out there whose livelihoods depend on understanding, interpreting, and reacting to what the data shows. Learn to read between the lines. [/QUOTE]

Verification: What is a Syracuse fan's favorite color?