Predictor and What-if — Live Proof of Strategies and a Journey Through Time

Imagine two traders. The first shows you a shiny table: "My strategy made +400% over five years." The second says: "Here's a list of stocks I picked this morning. Let's watch them together day by day and see how it turns out." Who do you trust more?

The first table is a backtest — it can be fine-tuned and embellished. The second list is live forward tracking — and that can't be faked. The future hasn't happened yet. This difference is the foundation of two tabs: Predictor (live tracking) and What-if (time simulator).

---

Part 1: Predictor — When Strategy Plays Live

What It Does

For each verified strategy, the system records which stocks the strategy has currently selected, and measures day by day how they perform. The list is "frozen" in time, and the yield is tracked forward. Technically, an out-of-sample test — the strategy runs on data that did not exist at the time of design.

★ Insight: A backtest answers "how would it have behaved in the past?". Predictor answers the more important question: "does it behave the same way in a future that no one could have known in advance?". The latter cannot be bent.

Why Live Tracking is the Toughest Test

A backtest has a weakness: the one who tunes it knows history in advance → overfitting, period selection, look-ahead bias. Live tracking has none of these levers.

★ Insight: "In-sample everything works, out-of-sample is where reputations die." Almost everything works in a backtest; only live does the truth reveal itself.

How to Read Live vs Backtest

Rule: when live data and backtest diverge, trust the live data. Example: the Champion strategy looked great in the backtest, but live it lagged by ~7 pp → honestly removed from the flagship position and replaced by Top Composite.

★ Insight: A good platform is not recognized by all strategies winning, but by admitting when one is losing.

Why Data Needs Time

To claim that a strategy has a real edge (not luck), you need a large enough sample. Short horizons (swing) mature in ~3 months, long ones (Top Composite, Multibagger) year+. Until it matures, you'll see "data collection in progress".

★ Insight: A small sample is the most common way to self-deception. Three wins in a row look like a genius system — often it's a coin toss.

---

Part 2: What-if — A Journey Through the Time Machine

What It Does

Answers "How much would I have today if I had back then…?". Two levels: pure price history up to 20 years (NVDA, $10,000, 10 years ago → value today) and strategy replay ~1 year (from score_history).

The Power of Compounding

After Years	Value from $10,000 (15%/year)
5 years	~ $20,100
10 years	~ $40,500
20 years	~ $163,700

Between the 15th and 20th year, more than $80,000 was added — more than the entire value in the 10th year. The biggest gains come at the end.

★ Insight: Beginners underestimate time and overestimate stock selection. The truth is the opposite: many years in a reasonable investment beat a few years in a genius tip.

The Big Trap: Survivorship Bias

When What-if shows you that NVDA turned $10,000 into hundreds of thousands, it's true about the past. Bad conclusion: "just find another NVDA". You see NVDA because it won. Next to it stood hundreds of "promising" stocks that went bankrupt and you don't see them today (they disappeared from the market). You're only looking at the winners and forgetting the graveyards of the losers.

★ Insight: A past winner does not tell you where the next one will be. It only says that winners exist. There's a chasm between "exceptions exist" and "I'll hit that exception". What-if is great for understanding compounding, but a lousy advisor for picking a stock for the future.

---

Part 3: How They Fit Together

   Predictor  ←——  STRATEGY  ——→  What-if
   (live proof,                  (time travel, compounding,
    out-of-sample)                BUT survivorship bias)

★ Insight: Trust in a strategy is not built on a single number. A backtest gives a hypothesis, What-if shows potential, Predictor confirms or refutes the hypothesis live. Only when all three show the same and the sample is large enough, do you have a solid foundation.

---

Practical Procedure

1. Start with the strategy, not the tool. Treat the backtest as a hypothesis.
2. Open Predictor and look at the age of the data (mature vs collecting).
3. Compare live vs backtest. Resolve discrepancies in favor of live data.
4. Use What-if for context and motivation, not for picking a tip.
5. Combine it all — trust = backtest makes sense + live data confirms on a large sample + you know the past is no guarantee.

---

Tips

1. Give Predictor time (short ~3 months, long year+).
2. When data fights, live wins.
3. What-if for compounding, not for prediction.
4. Follow multiple strategies simultaneously.
5. Write down expectations in advance (against hindsight rationalization).

---

Common Mistakes

Extrapolating past winners · ignoring that NVDA was an exception · premature judgment from a small sample · blind faith in backtest despite live data · mistaking illustration for advice.

---

Conclusion

Predictor provides the most precious thing — an unfalsifiable, time-stamped proof that a strategy works or not. What-if vividly shows the power of time and compounding — and if used honestly, protects you from the illusion that it's enough to hit the next big winner. Use both. But keep your head switched on.

---

QMA is an analytical and educational platform. Decisions are made at your own risk; consider consulting a licensed advisor.*