Turnover Difference off a Win/Loss
Note: Please read the Turnover Difference methodology basic articles first if you are unfamiliar with this approach.
The Turnover Difference has won over a twenty year span (and probably longer if we went back further in time!) at a remarkable rate, including a nifty 59% record from 2001 to 2003. Still, the tempatation with a method that produces so many plays is to see if any "tighteners" can be found that will improve the record even further.
There are dangers of course to this, but a reasonable course for some further investigation is to look at the outcome of the last game played by both the "turnover difference" selected team, and the opponent.
Since it's nice to have two data sets to test against (with the first you can see a long range view and make some hypotheses about the data, with the second you can retest your earlier conclusions), we will use a 1983-2000 set, and a 2001-2003 set.
Team off a WIN, Opponent off a WIN
| T/O Diff |
0.5-0.9 |
1.0-1.4 |
1.5-1.9 |
2.0-2.4 |
2.5+ Diff |
ALL |
| HomeFavs |
24 - 13 |
12 - 12 |
5 - 2 |
4 - 8 |
4 - 2 |
49 - 37 |
| HomeDogs |
11 - 9 |
14 - 10 |
9 - 5 |
5 - 10 |
4 - 6 |
43 - 40 |
| AwayFavs |
4 - 5 |
4 - 6 |
3 - 4 |
1 - 2 |
0 - 0 |
12 - 17 |
| AwayDogs |
24 - 19 |
16 - 24 |
13 - 8 |
9 - 1 |
1 - 4 |
63 - 56 |
|
| Favorites |
28 - 18 |
16 - 18 |
8 - 6 |
5 - 10 |
4 - 2 |
61 - 54 |
| Underdogs |
35 - 28 |
30 - 34 |
22 - 13 |
14 - 11 |
5 - 10 |
106 - 96 |
| Home Teams |
35 - 22 |
26 - 22 |
14 - 7 |
9 - 18 |
8 - 8 |
92 - 77 |
| Away Teams |
28 - 24 |
20 - 30 |
16 - 12 |
10 - 3 |
1 - 4 |
75 - 73 |
|
| ALL PICKS |
63 - 46 |
46 - 52 |
30 - 19 |
19 - 21 |
9 - 12 |
167 - 150 |
| WIN % |
58 % |
47 % |
61 % |
48 % |
43 % |
53 % |
| Spread Range |
7.5+ points |
3.5 to 7 |
0 to 3 |
| Favorites |
5 - 9 |
27 - 27 |
29 - 18 |
| Underdogs |
32 - 23 |
46 - 45 |
28 - 28 |
Comment: A much lower overall win percentage than the T/O Difference norm. Small favorites at 29-18 are the only 'spot plays' worth checking for the second data set test.
Team off a WIN, Opponent off a LOSS
| T/O Diff |
0.5-0.9 |
1.0-1.4 |
1.5-1.9 |
2.0-2.4 |
2.5+ Diff |
ALL |
| HomeFavs |
11 - 10 |
10 - 8 |
9 - 3 |
4 - 1 |
3 - 2 |
37 - 24 |
| HomeDogs |
6 - 5 |
6 - 4 |
1 - 1 |
1 - 2 |
1 - 1 |
15 - 13 |
| AwayFavs |
10 - 11 |
4 - 6 |
5 - 3 |
2 - 0 |
0 - 0 |
21 - 20 |
| AwayDogs |
23 - 14 |
15 - 11 |
3 - 6 |
3 - 4 |
6 - 5 |
50 - 40 |
|
| Favorites |
21 - 21 |
14 - 14 |
14 - 6 |
6 - 1 |
3 - 2 |
58 - 44 |
| Underdogs |
29 - 19 |
21 - 15 |
4 - 7 |
4 - 6 |
7 - 6 |
65 - 53 |
| Home Teams |
17 - 15 |
16 - 12 |
10 - 4 |
5 - 3 |
4 - 3 |
52 - 37 |
| Away Teams |
33 - 25 |
19 - 17 |
8 - 9 |
5 - 4 |
6 - 5 |
71 - 60 |
|
| ALL PICKS |
50 - 40 |
35 - 29 |
18 - 13 |
10 - 7 |
10 - 8 |
123 - 97 |
| WIN % |
56 % |
55 % |
58 % |
59 % |
56 % |
56 % |
| Spread Range |
7.5+ points |
3.5 to 7 |
0 to 3 |
| Favorites |
10 - 10 |
26 - 10 |
22 - 24 |
| Underdogs |
13 - 15 |
23 - 22 |
29 - 16 |
Comment: a little better results, but still below the total performance. It's at least a very consistent win rate across turnover difference level zones.
Team off a LOSS, Opponent off a WIN
| Diff --> |
0.5-0.9 |
1.0-1.4 |
1.4-1.9 |
2.0-2.4 |
2.5+ Pts |
ALL |
| HomeFavs |
20 - 17 |
22 - 9 |
10 - 5 |
1 - 5 |
3 - 3 |
56 - 39 |
| HomeDogs |
23 - 15 |
31 - 20 |
23 - 16 |
12 - 7 |
20 - 11 |
109 - 69 |
| AwayFavs |
2 - 1 |
3 - 0 |
3 - 1 |
2 - 1 |
1 - 0 |
11 - 3 |
| AwayDogs |
33 - 20 |
31 - 21 |
20 - 20 |
16 - 14 |
13 - 7 |
113 - 82 |
|
| Favorites |
22 - 18 |
25 - 9 |
13 - 6 |
3 - 6 |
4 - 3 |
67 - 42 |
| Underdogs |
56 - 35 |
62 - 41 |
43 - 36 |
28 - 21 |
33 - 18 |
222 - 151 |
| Home Teams |
43 - 32 |
53 - 29 |
33 - 21 |
13 - 12 |
23 - 14 |
165 - 108 |
| Away Teams |
35 - 21 |
34 - 21 |
23 - 21 |
18 - 15 |
14 - 7 |
124 - 85 |
|
| ALL PICKS |
78 - 53 |
87 - 50 |
56 - 42 |
31 - 27 |
37 - 21 |
289 - 193 |
| WIN % |
60 % |
64 % |
57 % |
53 % |
64 % |
60 % |
| Spread Range |
7.5+ points |
3.5 to 7 |
0 to 3 |
| Favorites |
6 - 6 |
20 - 14 |
41 - 22 |
| Underdogs |
82 - 57 |
90 - 63 |
50 - 31 |
Comment: Aha! The turnover difference selection, when coming from a loss versus an opponent off a win improves from the overall 57% expected cover rate to 60%. It's an even healthier 63% in games where the line is a field goal or less.
Team off a LOSS, Opponent off a LOSS
| Diff --> |
0.5-0.9 |
1.0-1.4 |
1.4-1.9 |
2.0-2.4 |
2.5+ Pts |
ALL |
| HomeFavs |
13 - 16 |
18 - 9 |
7 - 5 |
5 - 1 |
0 - 2 |
43 - 33 |
| HomeDogs |
14 - 14 |
10 - 8 |
5 - 3 |
5 - 2 |
1 - 2 |
35 - 29 |
| AwayFavs |
6 - 6 |
5 - 1 |
2 - 0 |
0 - 0 |
1 - 0 |
14 - 7 |
| AwayDogs |
25 - 25 |
30 - 17 |
11 - 10 |
3 - 8 |
8 - 3 |
77 - 63 |
|
| Favorites |
19 - 22 |
23 - 10 |
9 - 5 |
5 - 1 |
1 - 2 |
57 - 40 |
| Underdogs |
39 - 39 |
40 - 25 |
16 - 13 |
8 - 10 |
9 - 5 |
112 - 92 |
| Home Teams |
27 - 30 |
28 - 17 |
12 - 8 |
10 - 3 |
1 - 4 |
78 - 62 |
| Away Teams |
31 - 31 |
35 - 18 |
13 - 10 |
3 - 8 |
9 - 3 |
91 - 70 |
|
| ALL PICKS |
58 - 61 |
63 - 35 |
25 - 18 |
13 - 11 |
10 - 7 |
169 - 132 |
| WIN % |
49 % |
64 % |
58 % |
54 % |
59 % |
56 % |
| Spread Range |
7.5+ points |
3.5 to 7 |
0 to 3 |
| Favorites |
9 - 7 |
26 - 25 |
22 - 8 |
| Underdogs |
35 - 26 |
55 - 41 |
22 - 25 |
Comment: Back more in line with typical turnover difference averages, but curiously when you limit it to games with a 1.0+ or higher turnover difference margin, the numbers are fantastic -- 111-71 (61%), making it well worth looking for in the recent years...
The Second Look: 2001 to 2003
Rather than reproduce entire tables for the last three seasons, since there are few games the details are less significant, we'll go straight to the summary comparisons.
| TM |
Opp |
'83-00 |
Win% |
'01-03 |
Win% |
| Win |
Win |
167 - 150 |
53% |
41 - 25 |
62% |
| Win |
Loss |
123 - 97 |
56% |
25 - 25 |
50% |
| Loss |
Win |
289 - 193 |
60% |
26 - 24 |
52% |
| Loss |
Loss |
169 - 132 |
56% |
43 - 21 |
67% |
|
So, as is often the case, a second data set will yield different splits than was seen in the first set. In the table above we see the top '83-00 combo falls to a meager 52% in the recent timeframe, while the least successful pairing of the large data set was the second strongest the second time around, hitting 62%.
As for some of the other breakouts mentioned:
- In "Win-Win" cases where small favorites were 29-18 in the original test, they were 5-3 in the retest. Solid enough, but nothing too revealing.
- In "Loss-Win" games, where plays were 63% with lines of a field goal or less, the retest is showing a 10-6 mark, keeping in line with the original.
- In "Loss-Loss" games, with 1.0+ turnover difference margins being 61%, the numbers show 25-15 (63%) the second time out, but you are slicing off an 18-6 mark in the 0.5 - 0.9 range for the 2001-2003 season...
What to make of all this?
Turnover Difference selections coming off a loss are slightly better performers over the years,
and all three special situations remarked upon above held up fine over the second data set. However, the "two tales" did not agree very well on the two different timespans, which perhaps argues for simply following the turnover difference story 'as is' rather than believing that the most recent outcome for each team lends much extra to the equation.
|
