AXËL MOTORSPORT · THE RALLYCROSS DATABASE
RESULTS ARCHIVE

THE COMPETITIVE PICTURE

Who’s on the rise?

The front runners. The movers. The battles behind the numbers.

Statistical strength and uncertainty, or explore the legacy field score

STATISTICAL MODEL
CONNECTED DRIVERS122Ranked within this competition group
CLASS EVENTS84Within the selected event pool
MODELBradley–TerryOpponent adjusted · regularized
STABILITY CHECK64 refitsOptimization converged
Checked against later, unseen events

Train on the earliest 80% of event dates, test on the remaining 20%: 0.168 Brier score versus 0.250 for a 50/50 guess. Lower is better. 15 test events, 160 eligible non-tied pairs with known, connected drivers. This split beats the simple baseline.

Strength, with room for doubt

Dot = fitted score · line = 90% event-resampling range · dashed line = 1000 baseline

95010001450Tom HarringtonTom Harrington: 1306, 19 starts; resampling range 1254 to 13591306Shawna BowmanShawna Bowman: 1219, 33 starts; resampling range 1163 to 12611219Ryan FinchRyan Finch: 1202, 4 starts; provisional1202*Chris RaglinChris Raglin: 1181, 6 starts; resampling range 1128 to 12361181Samuel GesuelleSamuel Gesuelle: 1181, 5 starts; resampling range 1083 to 12371181Jason FullerJason Fuller: 1146, 3 starts; provisional1146*Evan MarkewyczEvan Markewycz: 1138, 7 starts; resampling range 1058 to 12111138Jacob SobiechJacob Sobiech: 1117, 5 starts; resampling range 1063 to 11851117Donald CarlDonald Carl: 1082, 1 starts; provisional1082*Dwight WoodDwight Wood: 1080, 1 starts; provisional1080*Ian SoderbomIan Soderbom: 1080, 2 starts; provisional1080*Hubert BorowskiHubert Borowski: 1073, 2 starts; provisional1073*

Overlapping ranges mean the apparent ordering is unstable. These measure sensitivity to the indexed events, not a guarantee of future performance. * Fewer than five rated starts.

Evidence-based standings

122 drivers · rank within selected group
RANKDRIVER / CARMODEL SCORE90% STABILITY RANGESTARTSOPPONENTS
113061254–13591942
212191163–12613364
31202Too little history418
411811128–1236622
511811083–123759
61146Too little history36
711381058–1211727
811171063–1185520
91082Too little history19
101080Too little history110
111080Too little history28
121073Too little history214
131072Too little history17
141071923–1157635
151068Too little history115
161064Too little history15
171059Too little history11
181059Too little history12
191058Too little history415
201054Too little history329
211053Too little history110
221050Too little history115
231049Too little history413
241046Too little history212
251045Too little history15
261042Too little history211
271039Too little history311
281039Too little history110
291039Too little history16
301038Too little history317
311036Too little history19
321035Too little history17
331034Too little history18
341032Too little history115
351029Too little history14
361026Too little history110
371024Too little history16
381023Too little history14
391023Too little history15
401022Too little history115
411022Too little history311
421021Too little history27
431017Too little history18
441016Too little history49
451016Too little history17
461016Too little history13
471014Too little history212
481014Too little history17
491013Too little history115
501013Too little history17
511012Too little history110
521008Too little history13
531008Too little history216
541005Too little history112
551004Too little history29
561004Too little history115
571002Too little history18
58998Too little history110
59996Too little history13
60995926–1064712
61995Too little history115
62994Too little history17
63992Too little history49
64992Too little history12
65991Too little history15
66989Too little history19
67989Too little history15
68989Too little history14
69988Too little history12
70988921–10502449
71986Too little history19
72986Too little history115
73984Too little history110
74982Too little history16
75980935–1023825
76978Too little history28
77975Too little history17
78972Too little history13
79972Too little history27
80969Too little history16
81969Too little history15
82967Too little history115
83966Too little history18
84966Too little history15
85964Too little history15
86962Too little history311
87962Too little history16
88958Too little history115
89958Too little history212
90958Too little history318
91958Too little history112
92955Too little history17
93955Too little history38
94955Too little history11
95952Too little history35
96949Too little history12
97949Too little history115
98944Too little history35
99944Too little history18
100940Too little history115
101937Too little history15
102936Too little history16
103935Too little history17
104933Too little history18
105930Too little history115
106929Too little history13
107928Too little history12
108928Too little history12
109925Too little history12
110924Too little history313
111921Too little history11
112917Too little history14
113915Too little history28
114914Too little history425
115912Too little history24
116912Too little history414
117902849–9461834
118889Too little history23
119881Too little history210
120878Too little history316
121867801–9281936
122734683–7911028
The science, assumptions and limitations

1. Compare like with like

Each class and event pool is modeled separately. Every scored class result becomes comparisons against the other classified drivers in that event. A win is 1, a tie is ½, a loss is 0. No-time entries, solo classes and duplicate driver entries within one class are excluded.

2. Estimate opponent-adjusted strength

We fit a Bradley–Terry model: P(A beats B) = 1 / (1 + exp(θB − θA)). Our implementation minimizes weighted logistic loss plus ½Σθ². Each comparison weighs 1/(field size − 1), so a driver contributes one unit per event. Regularization pulls short or perfect records toward the baseline. Display score = 1000 + (400 / ln 10) × θ; a 400-point gap means 10:1 fitted pairwise odds. This is separate from F1 points and the legacy field score.

3. Show uncertainty and test forward

We refit 64 times after resampling whole events with replacement and show the middle 90% of scores. Resampling keeps opponents from one race together. Missing drivers in a resample return to the baseline. These are stability ranges, not Bayesian credible intervals; sparse records can remain poorly identified. We also fit on the earliest 80% of event dates and evaluate later results using weighted Brier loss. Test results cover known, connected drivers only and can vary with this single split. Parameters were fixed before this check.

4. Respect the gaps

Drivers without a chain of shared opponents form separate competition groups; we do not predict between them. The model assumes stable ability across its selected history and cannot observe practice, weather, tires, mechanical failures, or unindexed events. Five starts is a display threshold, not proof of reliability. The ranking is an experimental statistical estimate, not an official SCCA championship or an assessment of a person.

5. Cars and win probabilities

The optional driver–car model adds a strongly regularized car term (penalty 2Σβ²). A car label needs at least eight entries and three drivers before receiving an adjustment; otherwise it adds zero. Labels remove the year and normalize punctuation, but do not resolve every trim or spelling. Shared labels are not guaranteed identical equipment. Driver choice and car preparation are confounded: a positive term does not prove the car caused better results.

Field win chances use the Plackett–Luce choice formula, exp(θdriver + βcar) / Σfield exp(θdriver + βcar). It is a scenario conditional on this field and all entries posting comparable scored results. It excludes breakdown/DNF risk, is not calibrated race-winning odds, and must sum to 100% before rounding. Adding a competitor changes everyone’s chance. The car model’s holdout is reported separately.