How to Read the Portfolio Chart
June 16, 2026 · 7 min read
The Portfolio Chart is QuantBuilder's primary performance view. It shows how one or more combined models performed over a selected time window — confidence levels, actual returns by day, how many picks were generated, and a running equity curve. If you've seen this chart on social media and weren't sure what you were looking at, this guide explains every element.
Portfolio Chart — Consumer Cyclical combined model, 120-day view
Three Panels, One Picture
The chart is stacked into three panels. Reading all three together is how you understand what a model is actually doing — not just the final return number, but why, when, and how it got there.
Top Panel
Confidence + Returns
How confident was the model each day, and what did those picks actually return?
Middle Panel
Picks per Day
How many picks did the model generate each day? This is the most important signal.
Bottom Panel
Equity Curve
The cumulative return of all model picks vs. the benchmark, using a fixed position size per pick.
Confidence + Returns
The top panel plots two things on the same timeline: how confident the model was on each day (the dots), and what the picks from that day actually returned (the bars).
Green dots — Daily Confidence %
Each dot represents the average confidence score across all picks generated on that day. Confidence is the model's internal estimate, calibrated from thousands of historical examples, of how strongly the current setup matches what it learned to find. A day showing 58% confidence means every pick that day had conditions the model recognizes — not a guess, but a pattern match.
Teal bars (above zero) — Positive Return Days
When a bar extends above the zero line in teal, the picks from that day returned a positive result over the model's target holding window. The height of the bar shows the magnitude. A 5% bar means the model's picks, averaged across the holding period, returned +5%.
Red bars (below zero) — Negative Return Days
Red bars below the line are days where picks lost ground over the holding window. Having negative return days is normal and expected — what matters is the balance between red and teal bars over the full period, and the asymmetry between the size of winners and losers.
Yellow dots — Today's picks (no return yet)
Yellow dots mark the most recent picks that don't yet have a completed return window. The model generated picks on those days, but the holding period hasn't expired — so there's nothing to show as a bar yet. This is expected for any date within the last few days.
The bar label says: "Each bar = total return of picks over model's target window." If you built a 5-day model, Day 1 might be +6%, Day 2 +8%, Day 3 −5% — the bar shows the cumulative result at the end of the 5-day window, not a single-day move.
Picks per Day — The Most Important Signal
The middle panel shows how many picks the model (or combined models) generated on each day. Long picks are shown in green; short picks in red.
This is the single most diagnostic element of the entire chart. The pick count is not random noise — it is the model's opinion about current market conditions. When it sees setups that match what it learned to find in training, it generates picks. When conditions don't match, it barely picks. It doesn't force itself to find setups that aren't there.
Three distinct phases are visible in this example:
The spike — 43 picks in a day (yellow box)
In early April, both models fired simultaneously: 43 picks in a single session, then 43 again the next day, then 23. This was the Consumer Cyclical sector rotation — after months of the models staying quiet, macro conditions shifted, sector valuations reset, and suddenly dozens of stocks matched the model's learned patterns all at once. The pick count spike is the early rotation signal, visible before any sector headline explains it.
The quiet period (blue box)
During the blue-highlighted stretch, the model had 1–3 picks per day, sometimes none. This is the model saying: "the sector conditions don't match my training right now." It's not broken. It's not confused. It's waiting. Look at the equity curve during this period — it goes flat. Because if you're barely in the market, you can't lose much. The model protects you by staying quiet.
Recent activity (red box)
The red-outlined section shows the most recent days, typically 10–18 picks per day. The model has found conditions it recognizes and is actively generating picks. These are the days closest to "today" and most relevant to current trading decisions.
A model that generates zero picks on a given day is performing correctly — it simply doesn't see the setup it was trained to find. A model that generates picks in bad conditions would be worse, not better. The quiet periods are a feature, not a failure.
Equity Curve
The equity curve shows cumulative return over the period. Two lines:
Blue line — Model equity curve
The hypothetical return if you allocated a fixed percentage of your portfolio per pick (1%, 2%, or 3% — switchable with the buttons at top left of this panel). Every closed pick's return is added to the running total. This is not a backtest — it's the actual live performance of every pick the model generated during the selected period.
Yellow line — Benchmark
The selected benchmark (Consumer Cyclical in this example) over the same period. This lets you see whether the model is actually generating alpha above the sector it operates in, or just riding the sector move.
Notice the shape of the blue line. It stays nearly flat through the quiet period — the arrows in the chart point to exactly this. Then it rises during the rotation, matching the spike in pick count. The equity curve doesn't just tell you the final return; it tells you when the model was active and when it was standing aside.
In this example, the model finishes up while the Consumer Cyclical benchmark is down on the full period. The model didn't follow the sector blindly — it waited for conditions to align, deployed when the rotation arrived, and avoided most of the sector's drawdown.
Reading the Bottom Stats Bar
Below the equity curve, four numbers summarize the full period:
Days Analyzed
84
Trading days in the selected period where the model had at least one pick
Avg Confidence
48.0%
Average model confidence across all picks in the period
Avg Return
+0.45%
Average return per pick, using BUY OPEN entry with -5% stop loss
Win Rate
44%
Percentage of picks that closed positive before stop loss
A 44% win rate with a positive average return is viable because the wins are larger than the losses on average. The stop loss at −5% caps the downside on each pick, while winners run to the model's target window. This asymmetry — small controlled losses, larger gains on winners — is the foundation of a positive-expectancy system even at sub-50% win rates.
The Control Bar
The row of controls above the chart lets you change what you're measuring without rebuilding anything.
PERIOD — 14d / 30d / 90d / 120d / 180d
How far back to look. The 120-day view shown here captures the full Consumer Cyclical rotation story — the long quiet period, the spike, the sustained pick activity. Shorter windows (14d, 30d) show the most recent behavior; longer windows (180d) reveal how the model performed across different market regimes.
SL — Stop Loss %
The stop loss applied to every pick when calculating returns. Changing this recalculates all the bars and stats instantly. At −5%, a pick that moves down 5% from entry is counted as a −5% loss and closed. At −10%, the same pick has more room to breathe — potentially recovers, potentially loses more. The chart shows you how sensitive the model's results are to stop loss settings.
× TOP
Filters to show only the top-ranked picks per day (highest confidence). Instead of evaluating all picks, you see how the model performs if you only take the best ones.
⊘ OVERLAP
When combining multiple models, OVERLAP filters to only show picks that appear in two or more models simultaneously. These are the highest-conviction picks — multiple models agreeing on the same stock on the same day. Often a smaller list with better hit rates.
⊗ CONFLICT
Hides picks where one model goes long and another goes short on the same stock. Conflicting signals on the same ticker are a reason to skip that pick; this toggle removes them from the view automatically.
ENTRY — BUY OPEN
Entry method for return calculations. BUY OPEN uses the next morning's opening price as the entry point. This is the realistic entry for a model that generates picks overnight — you see the pick before the open, then enter at open.
BENCHMARK
Choose what to compare the equity curve against — Consumer Cyclical, S&P 500, NASDAQ, or other sector indexes. The benchmark is shown as the yellow line in the equity curve panel.
What This Chart Actually Tells You
The portfolio chart answers three questions that raw return numbers can't:
- 1.
Is the model consistent, or did one lucky week drive all the returns? The bars and equity curve tell you. A smooth rising curve with evenly distributed positive bars is very different from a flat line punctuated by one big week.
- 2.
Is the model using market conditions intelligently? The picks-per-day panel tells you. A model that fires 40 picks in a down market isn't disciplined. A model that stays quiet and then fires when conditions align is doing exactly what it should.
- 3.
Is the model actually better than just owning the sector? The benchmark comparison on the equity curve answers this directly. If the blue line finishes above the yellow line, the model added value above passive sector exposure. If it finishes below, it didn't.
In this Consumer Cyclical example, all three answer favorably: returns are distributed across the full period, the model stayed quiet through the sector drawdown and deployed only when conditions aligned, and the equity curve finished above the benchmark while the sector itself was still in the red.
BUILD YOUR FIRST MODEL
Train a machine learning stock picking model in minutes — no code required. Your picks generate every morning before the market opens.