Hello people of the ISFL and DSFL! On the 26th of March, the Season Prediction Point Task was posted in the forums, as it is before the start of every season. Me and two other users, who have not had their data used in this study, made 3 separate tables that went through the first entries of the prediction form. After these tables were posted in different chats, however, a discussion arose. Many users were contesting the data of these tables. This doubt did not arise from the tables and the calculations themselves, but from the users who submitted themselves. One user went as far as saying “I would also keep in mind that the people in this league can be dumb” (referring only to other people’s capacity to correctly predict/assess the situation of different ISFL teams. Are the people in this league really dumb? This is what this study seeks to find out.
In it, I went over the data of the first 50 respondents to the Season Prediction Point Task in S26 and S27 and compared their average answers to the actual standings of those seasons. The objective was to find out how much the variation was between those first respondents and the actual results and how correctly do ISFL users make their predictions. I took only the first 50 respondents into consideration since that was the number I used for my S28 season prediction, since only 50 people had responded to it at the time of the making of the table. I also did not include predictions of playoff champions, as I believe that was tougher to establish a mathematical pattern that could contribute to this study.
I inputted all 50 first season predictions of Seasons 26 and 27 into a table and calculated the average standing of each team according to what users predicted. Then, I calculated what I call the “average deviation”, or AD for short, which is how much users deviated from a team’s real position in the standings in their predictions. It can be algebraically expressed as AD = |P – S|, (P is the Average Prediction of a team and S is the Season Standings of said team). And finally, I summed all AD’s per team and divided by 7 (the number of teams in each ISFL conference, since teams are ranked on a scale of 1 to 7) to find out the total deviance: how much do people deviate from reality when predicting future standings.
The results were surprising. The discussion I read made me believe people were generally wrong, as it would be expected when predicting future events that may be altered by an infinite number of circumstances. What followed, however, neither disproved it nor did it prove it.
In Season 26, total deviation was 3.03, which means that total deviation percentage was 22%, and, consecutively, people got 43% of their predictions right. In Season 27, results were very similar. Total deviation was 3.98 and total deviation percentage was 57%, meaning 72% of predictions were correct. Averaging the two seasons that were analyzed, total deviation was 3.505 and total deviation percentage was 50%, which means half of all 1400 predictions analyzed were correct.
Does that mean that other people’s predictions are to be taken into consideration for you to make your own? That is up to debate.
If you are a new user in the league, as I am, 50% could be considered a great percentage, especially since guessing teams randomly grants only a 14% chance. It is also important to highlight that ASFC has been a volatile league over the past few seasons, and ASFC AD is 2.5 (35%), very high compared to NSFC’s AD of 1 (15%), so maybe other people aren’t to be trusted when making ASFC predictions but can be trusted to predict the NSFC.
If you are a veteran user, however, maybe it is better to just stick to your guts. When David Glidden, from towarddatascience.com, decided to use an algorithm to gather predictions from 5 experts/expert sites to guess the results of NFL games, these experts had a success rate of 66.9% on average, and so did he at the end of the 2015 season. While it may be stretch to consider yourself a ISFL “expert”, veteran users are the closest to an ISFL expert that there are, especially if one is a GM or has another role of importance in the league. So predictions may not help you too much, or even bring you down.
TL;DR: I calculated how good people were at predicting the ISFL standings over the last two seasons. The results showed that over Season 26 and 27, half (50%) of all 1400 predictions were correct. I concluded that:
• If you are a new user of the ISFL, using predictions could provide a slight advantage.
• If you are a veteran user of the ISFL, sticking to your guts is probably the best option.
• If one is to use user’s predictions to determine their own, users are twice as right guessing the standings of the NSFC as they are of the ASFC, so follow what people are saying about the NSFC, and maybe decide the ASFC on your own.
I would also like to shoutout @slate, who made a table calculating the average prediction of the first 40ish predictions of the S28 standings, which was an inspiration to this study.
In it, I went over the data of the first 50 respondents to the Season Prediction Point Task in S26 and S27 and compared their average answers to the actual standings of those seasons. The objective was to find out how much the variation was between those first respondents and the actual results and how correctly do ISFL users make their predictions. I took only the first 50 respondents into consideration since that was the number I used for my S28 season prediction, since only 50 people had responded to it at the time of the making of the table. I also did not include predictions of playoff champions, as I believe that was tougher to establish a mathematical pattern that could contribute to this study.
I inputted all 50 first season predictions of Seasons 26 and 27 into a table and calculated the average standing of each team according to what users predicted. Then, I calculated what I call the “average deviation”, or AD for short, which is how much users deviated from a team’s real position in the standings in their predictions. It can be algebraically expressed as AD = |P – S|, (P is the Average Prediction of a team and S is the Season Standings of said team). And finally, I summed all AD’s per team and divided by 7 (the number of teams in each ISFL conference, since teams are ranked on a scale of 1 to 7) to find out the total deviance: how much do people deviate from reality when predicting future standings.
The results were surprising. The discussion I read made me believe people were generally wrong, as it would be expected when predicting future events that may be altered by an infinite number of circumstances. What followed, however, neither disproved it nor did it prove it.
In Season 26, total deviation was 3.03, which means that total deviation percentage was 22%, and, consecutively, people got 43% of their predictions right. In Season 27, results were very similar. Total deviation was 3.98 and total deviation percentage was 57%, meaning 72% of predictions were correct. Averaging the two seasons that were analyzed, total deviation was 3.505 and total deviation percentage was 50%, which means half of all 1400 predictions analyzed were correct.
Does that mean that other people’s predictions are to be taken into consideration for you to make your own? That is up to debate.
If you are a new user in the league, as I am, 50% could be considered a great percentage, especially since guessing teams randomly grants only a 14% chance. It is also important to highlight that ASFC has been a volatile league over the past few seasons, and ASFC AD is 2.5 (35%), very high compared to NSFC’s AD of 1 (15%), so maybe other people aren’t to be trusted when making ASFC predictions but can be trusted to predict the NSFC.
If you are a veteran user, however, maybe it is better to just stick to your guts. When David Glidden, from towarddatascience.com, decided to use an algorithm to gather predictions from 5 experts/expert sites to guess the results of NFL games, these experts had a success rate of 66.9% on average, and so did he at the end of the 2015 season. While it may be stretch to consider yourself a ISFL “expert”, veteran users are the closest to an ISFL expert that there are, especially if one is a GM or has another role of importance in the league. So predictions may not help you too much, or even bring you down.
TL;DR: I calculated how good people were at predicting the ISFL standings over the last two seasons. The results showed that over Season 26 and 27, half (50%) of all 1400 predictions were correct. I concluded that:
• If you are a new user of the ISFL, using predictions could provide a slight advantage.
• If you are a veteran user of the ISFL, sticking to your guts is probably the best option.
• If one is to use user’s predictions to determine their own, users are twice as right guessing the standings of the NSFC as they are of the ASFC, so follow what people are saying about the NSFC, and maybe decide the ASFC on your own.
I would also like to shoutout @slate, who made a table calculating the average prediction of the first 40ish predictions of the S28 standings, which was an inspiration to this study.