The Banister model: how TriPaced sees your fitness in real time.
Most "AI coaches" treat fitness as a black box. TriPaced uses the same impulse-response model that pro coaches have leaned on for 30 years — and shows it to you, daily, with the actual numbers driving every plan change.
What is the Banister model?
In 1975, Eric Banister and colleagues at Simon Fraser University described training as an impulse-response system. Every session is a stimulus that produces two effects at once: fitness, a slow-building, slow-decaying adaptation, and fatigue, a faster-building, faster-decaying cost. Modelled performance is roughly fitness minus a weighted fatigue. When fatigue clears faster than fitness fades, you peak.
Fifty years later that idea still underpins how serious endurance training is measured. TriPaced expresses it with three numbers, computed from your training stress score (TSS) history:
- CTL — Chronic Training Load
- A rolling 42-day exponentially-weighted average of daily TSS. Your fitness proxy. It moves slowly, on purpose.
- ATL — Acute Training Load
- A rolling 7-day exponentially-weighted average of daily TSS. Your fatigue proxy. It reacts fast to a hard week or a missed one.
- TSB — Training Stress Balance
- CTL minus ATL. Negative means fatigued and loading; positive means freshening. The race-day target is roughly +5 to +15.
Why we expose the live model
Other AI coaches "decide" your week without showing the math. If you do not know why Tuesday turned into an easy spin, you cannot trust the system — and you cannot tell when it is wrong.
TriPaced shows CTL, ATL and TSB on every dashboard. Every plan change carries the rule that fired next to it: "TSB below −25, recovery activated", or "CTL ramp would exceed 8 per week, long ride capped". You are never asked to take the plan on faith.
Since June 2026, the engine also reads your load per discipline — swim CTL, bike CTL, run CTL — instead of a single averaged number. Most platforms still blend the three sports into one figure. We do not, because the legs that absorb a 100-TSS run are not the system that absorbs a 100-TSS ride.
What this means for you
You can argue with the model. When its number disagrees with how your body feels, you say so, and if you are right, the plan updates. A black box gives you nothing to argue with.
You can read your own trend. A CTL gain of around +20 across a 16-week build is realistic. Much more than that is a red flag, not a brag — and now you can see it happening instead of finding out on race day.
You can understand your taper. The final week looks easy because the math pushed TSB from roughly −18 to +8 in seven days, not because someone decided race week should be gentle.
The limits of Banister (the honest part)
The model is a tool, not an oracle. It does not see technical specificity, heat, altitude, fueling, illness or sleep debt. TriPaced layers those on top through completion feedback and per-discipline balancing — see multi-discipline load.
Banister also assumes TSS is a valid intensity proxy. For cycling power that holds well; for swim and strength it is an approximation, so we expose an override rather than pretend the number is gospel.
And the model can drift if every session is logged but none is logged honestly. Trust is not blind faith. Calibrate it with an FTP or threshold test roughly every six weeks, and the three numbers stay anchored to reality.
Frequently asked
What is a healthy CTL ramp rate for an Ironman build?
Around 5 to 8 TSS points per week. Above 10 is the red zone where most age-group injuries occur. TriPaced enforces this as a hard guard-rail — the planner will not generate a week that ramps faster than the limit.
My CTL is 80 today and the plan wants it at 110 by race day. Is that realistic?
A 30-point gain at a safe 5 to 8 points per week needs at least 4 to 6 weeks of clean build. If your race date does not leave that room, TriPaced tells you up front instead of quietly prescribing an injury-rate ramp.
Why does my TSB swing so much when I miss a session?
Because ATL has a roughly 7-day half-life. A missed long ride drops ATL fast, which pushes TSB positive, so the engine reads "freshening" and may re-prescribe intensity. Honest completion logging is the fix.
What is a good TSB for race day?
Plus 5 to plus 15 for a full Ironman, plus 10 to plus 20 for a 70.3. A negative TSB on race day means you under-tapered.
Does CTL transfer between swim, bike and run?
Only partially. Aerobic base transfers more than sport-specific neuromuscular endurance, which is exactly why TriPaced tracks per-discipline CTL.
Why does TriPaced's CTL differ from TrainingPeaks or intervals.icu?
Same math, different completion sources. Reconcile by importing your full training history so every platform sees the same sessions.
Keep reading
→ Multi-discipline load: why we don't average TSS across sports
→ Ask Claude "why did my TSB drop last week?" — the TriPaced MCP server
TriPaced builds a 16-week adaptive long-course plan around your real CTL, ATL and TSB — and shows you the numbers behind every change. The beta is open and we review every application by hand.
Apply for TriPaced beta →Not ready to apply? The beta is free and open — see your real CTL, ATL and TSB, with the reasoning shown behind every change.
See the glass box for yourself →Sources. Banister, E.W. (1975). A systems model of training for athletic performance. Australian Journal of Sports Medicine. · Coggan, A. & Allen, H. (2019). Training and Racing with a Power Meter, 3rd ed., VeloPress. · McGregor, S.J. et al. (2009). Modelling the effect of training on performance. · Foster, C. (1998). Monitoring training in athletes with reference to overtraining syndrome. Medicine & Science in Sports & Exercise.