← Back to the articles index

Winning time guessing

As prediction games got popular, racers are not only guessing on the winner and the top10, but often they also guess on the winning time on the current track. However I often felt these guesses are mostly just numbers based on racer's intuition. No, that's not a bad thing, but why not use some more accurate method? Well, here is mine.

Of course everyone knows it's harder to guess at the beginning of the month, than later. This article will also prove this. But what to choose as the base of these calculations? Leading time was the perfect choice, as it is availble thorughout the whole month, and is well documented.

So what did I do? I made a chart of the leading times on each track finished in the 2005 season, writing up only the current leader time each day at midnight. Then I divided these leading times with the final winner's time of that track. This way I got a percentage for each day, which is somewhere above 100%, and shows how that day's leading time relates to the final winning time. I have also made a nice graph for you:

I was delighted to see that except one track, ZCT49 the graphs were very similar. Why ZCT49 is so different? There was a very intense fight for the LTB, so racers sent teplays through the whole month, always making it only some milliseconds better. This kept them from maxing out the replay, as they were not searching for new tricks, but tuning the present ones. Intresting, isn't it? Seems like fighting hard for the LTB wont't give you a decent time at the end of the month.

But back to our topic. As you can see in the first days the leader's time is around 110-120% of the final winner's. Notice the strong drop in the graphs, as racers send stronger and stronger times thorughout the first racing week. Then most of the graphs get very linear and boring. This is the 2nd and 3rd week, where there is already a hard leading time, and people stop fighting. Then there are occassionally again some drops at the start of the 4th week, this is the last attack in order to earn some LTB. And finally on the last 2-3 days there is a strong drop - that's the effect of the quiet days, as everyone hides his best :)

So thats a nice graph, but how can we guess on the winning time from it? Well simple. I've averaged the percentages for each day. For example if its its 10th day of the month, you check the chart below, and see 109.4% Then you divide the current leading time with 1.094, and get a result close to the winning time. Possibly.

Why isn't this accurate? Well Can you see the Standard devitaton column? I won't explain it, if you have learned statistics you know what that is. These number shows the average difference from the averaged guess,in which interval the final time will be. Why is it so high? Because I calculated with only 6 tracks as a base for the results. For a considerably accurate result I should have done this with 30 tracks.

Maybe someday I'll. But until then, winning time guessing got slightly more scientifical :)

Zak McKracken, 2005.07.02

Submitted by Zak McKracken on 2005-07-02.