KiqIQ is built on a transparent analytical framework: Poisson distribution over xG inputs, explicit tournament-mode and venue adjustments, EV-based decision rules, and conservative Kelly staking. This page documents what KiqIQ does, what it doesn't do, and the limitations of the framework — so you can question it, recompute it, and decide for yourself.
Every fair-odds calculation in KiqIQ ultimately comes from a Poisson distribution over expected goals (xG). The model takes home and away xG inputs (typically 5-game rolling averages calibrated by league baseline), generates a scoreline distribution from 0-0 to 5-5, and sums that distribution into 1X2, BTTS, Over/Under, and AH probabilities. Poisson is well-established as a baseline football model and is more transparent than ML models — every output traces back to two xG inputs and a home advantage multiplier.
→ Transparent over predictive: a Poisson output can be questioned and recomputed by hand. ML predictions can't.
KiqIQ does not produce its own event data. xG inputs are pulled from public sources: FBref (Opta-derived), Understat, league-average baselines for less-covered competitions. Users feed these into the Poisson calculator manually. KiqIQ's value is the analytical workflow on top of public data — not the data itself.
→ Honest framing: users should know that the data quality of any KiqIQ output depends on the xG source they used as input.
For international tournaments and competitions with venue-specific factors, KiqIQ applies explicit calibration: ×0.85–0.90 xG discount for tournament-mode caution, ×0.85 visitor xG penalty at high-altitude venues (Mexico City, La Paz, Quito), +0.4 xG for tournament hosts, climate adjustments for extreme heat venues. Each adjustment is documented in the relevant guide and is applied by the user — KiqIQ does not auto-apply these because the user has more context about specific fixtures.
→ Bookmakers under-adjust for tournament-mode and venue factors. Explicit adjustment is the source of systematic edges.
KiqIQ doesn't recommend bets — it computes EV. EV = (P_model × decimal_odds) − 1. If EV is positive, the model thinks the bet has value. If EV is negative, the bet is house-edge. The decision to actually bet remains with the user. KiqIQ shows fair odds, EV percentage, and Kelly stake size, but does not press buy/sell buttons or push specific bets.
→ EV is testable: log every bet, compare actual return to predicted return over a large sample. Tipsters' "value" claims are not.
Kelly Criterion is the mathematically optimal stake-sizing method when EV is positive — but full Kelly produces extreme variance. KiqIQ defaults to quarter-Kelly (×0.25 of full Kelly) and recommends additional reductions for high-variance bets: ×0.20 for tournament fixtures, ×0.175 for "to qualify" markets with shootout exposure, ×0.0625 for same-game parlays. The conservative defaults reduce theoretical EV but keep the bankroll alive through losing streaks.
→ Variance kills bankrolls before EV kills bookmakers. Conservative staking is what makes positive EV survivable.
No model is right about every match. KiqIQ's limitations: (1) Poisson assumes goal independence — it doesn't capture team-news shocks, red cards, weather. (2) Public xG data lags the live market by hours. (3) Tournament samples are small (one fixture decides everything in knockouts). (4) The Poisson + EV framework works best in mid-tier league fixtures and is weakest in cup ties, derbies, and pre-determined fixtures where motivation overrides form.
→ A model that claims certainty is not a model. Real probability work is about quantifying uncertainty, not eliminating it.
KiqIQ doesn't publish "today's top tips" or claim a winning record. Tipsters by their nature select for sample bias — they remember winners and forget losers, and the maths usually doesn't survive a properly tracked sample.
KiqIQ doesn't use opaque deep-learning models to generate match probabilities. The probability engine stays Poisson — transparent, reproducible, debuggable. The AI layer sits on top: it explains the bet, stress-tests assumptions, and runs what-if scenarios — but never replaces the math underneath.
Recent results matter only insofar as they reflect underlying xG. A team on a 5-match winning streak with low xG is regressing. A team on a 5-match losing streak with high xG is undervalued. KiqIQ uses xG, not raw results.
KiqIQ does not promise profit. Positive EV is a long-term concept that requires hundreds of bets to manifest, not dozens. Disciplined bankroll management with conservative Kelly is the path. Ruined bankrolls aren't — even with positive EV.
The Poisson + EV framework works best in some contexts and worst in others. Be explicit about confidence before staking.
| Confidence level | Applies when | Use for |
|---|---|---|
| High confidence | Mid-tier domestic league, both teams' rolling 5-game xG samples > 60 minutes, no known team-news disruption, market has been settled for 2+ days | Standard-Kelly staking against the model |
| Moderate confidence | International tournament group stage, late-stage knockout, or domestic cup. Some tactical/motivation uncertainty. | Quarter-Kelly with 20-30% reduction |
| Low confidence | Knockout single-leg ties between unfamiliar opponents, derbies with motivation overrides, fixtures with confirmed major team-news disruption | Skip the bet or use minimum-stake exploration only |
| No confidence (do not bet) | Already-qualified or already-eliminated fixtures, post-trophy dead rubbers, friendlies, fixtures with insufficient xG data | No bet. The bookmaker margin is the only certain outcome. |
Real probability work is about quantifying uncertainty, not eliminating it. KiqIQ's outputs are probabilities — not predictions. A 65% home win probability is correct on average if 65 of every 100 such fixtures end in home wins. The 35 losses are not "model failures" — they are the model working as designed.
The discipline is to bet only when EV is positive, stake size is appropriate (quarter-Kelly or less), and the confidence level matches the situation. Over hundreds of bets at +5% EV with disciplined sizing, the bankroll grows. Over a dozen bets, anything can happen.
The EV calculation, fair-odds derivation, and EV vs implied-probability comparison framework.
Quarter-Kelly defaults, edge thresholds, variance discount for high-volatility bets.
Why even profitable bettors face deep drawdowns — what timescales actually matter.
The full Poisson workflow from xG inputs to probability output to value comparison.
Bankroll discipline, deposit limits, self-exclusion options, and warning signs to watch for.
The platform, the editorial principles, and the team philosophy behind KiqIQ.
Use the Poisson calculator to derive scoreline probabilities, then apply the EV framework to compare against bookmaker odds.
