Rule: Vercel pages are read-only. Private controls live on Railway only.
The big idea
Think of ScholarTrade like a “smart lemonade stand”
Imagine you own a lemonade stand. Your goal is to make a little profit each day.
But you also want to keep your money safe so you don’t go broke.
ScholarTrade is like a robot helper that:
Buys and sells crypto a bit like buying ingredients and selling lemonade.
Tries to make profit (so scholarships can be paid).
Has a strict safety guard that stops it taking silly risks.
Shows its work on public pages, so you can see what it’s doing.
Important: There are two “places” you look:
Public pages (Vercel) = the shop window. Anyone can look. No buttons that can trade.
Private ops (Railway) = the control room. Only you + your sons can log in and press buttons.
What you’ll see
What the dashboards mean (without finance-speak)
1
Equity = “How much money we have in total right now.”
If equity goes up, you’re winning. If it goes down, you’re losing.
2
Daily PnL = “How much we’ve won or lost today.”
Useful for spotting a bad day early.
3
Drawdown = “How far we’ve fallen from the best point recently.”
Like: if your high score was 100 and now you’re at 90, drawdown is 10%.
4
Positions = “Things we currently hold.”
If you see a position, the bot is in a trade.
How trading decisions happen
Three ‘brains’ + one ‘bodyguard’
ScholarTrade can run more than one strategy (brain) at once, but it has rules so they don’t fight.
DCA brain: buys in small pieces over time instead of “all at once” (like buying ingredients slowly so you don’t overpay).
Grid brain: likes to buy/sell in steps when price wiggles (like restocking lemonade when it’s cheap).
Trend brain: tries to catch bigger moves (like selling more when a big crowd shows up).
Risk bodyguard: can say “NO” and block trades if it thinks it’s too risky.
New: the “Meta brain” (head coach)
Think of this as a head coach watching all the brains.
The coach doesn’t invent trades; it decides which brain is allowed to drive right now.
Shadow mode = scouts with notebooks: every strategy still produces signals, but most of them are treated like “practice plays.” We record them and see how they would have done.
Gate = only one pilot flies the plane: for each coin (symbol), only the selected strategy is allowed to place live orders. Others can suggest, but they don’t touch the controls.
Cooldown = don’t change captains mid-storm: even if the ranking changes slightly, we don’t switch instantly. Switching is slow and conservative.
Fail-open safety: if the meta brain has no data yet (or errors), the bot won’t “freeze.” It will behave normally until evidence accumulates.
Where to see it: the Ops Console shows /api/meta/status + /api/meta/scoreboard.
The Public Transparency page shows a simplified meta status panel.
About the “Liquidity Reversal” (LR) strategy:
LR is experimental and is typically rolled out in stages.
It may be enabled for testnet forward-testing, while remaining disabled by default in production
until it has proven itself.
Why? Because “research code” and “live-trading code” have different standards:
we still need (1) forward-testing (shadow/paper), (2) production-grade guards (risk sizing, circuit breakers, kill-switch paths),
(3) operator observability (clear strategy attribution, rejects, and health signals), and (4) an explicit rollout toggle.
Safe rollout path: Backtest -> Shadow mode (log-only) -> Paper -> Small-size testnet -> Full enable.
If you want the nerdy details, see: docs/LIQUIDITY_REVERSAL_STRATEGY_EVALUATION.md
Why you might see multiple positions:
The bot can hold positions in different coins at the same time (e.g. BTC and ETH).
It also prevents two strategies from spamming the same symbol at once.
Strategies (simple, but complete)
What each strategy actually does
The bot can run multiple strategies at the same time. Each one has its own entry rules, sizing rules, and exits.
A “symbol ownership” rule prevents two strategies from fighting over the same coin.
1) DCA (Dollar Cost Averaging)
What it’s good at: slowly building a position without trying to “time the perfect bottom.”
When it buys: on a schedule (e.g. every 24h) and/or when price dips versus recent highs.
How it sizes: a fixed USD budget per cycle split by allocation (BTC vs ETH), with a dip multiplier that increases size on deeper drops.
How it takes profit: the engine can place exits (take-profit / stop-loss / trailing), or it can close based on strategy exit signals depending on the profile.
2) Grid
What it’s good at: choppy, sideways markets where price “wiggles.”
When it buys: at pre-set spacing below a moving center price.
How it sizes: usually a fixed USD order_size converted to quantity at the current price.
How it takes profit: either by selling at higher grid levels, or (in simplified modes) by letting the engine handle take-profit / stops.
3) Trend
What it’s good at: catching bigger directional moves.
When it buys: when indicators align (trend direction + strength + momentum + volume confirmation).
How it sizes: a fraction of equity (e.g. 2%–5% of the account value) converted to quantity.
How it takes profit: volatility-aware stops (ATR-based) + optional trailing stops to protect gains while letting winners run.
4) Liquidity Reversal (LR)
What it’s good at: sharp reversals after a “sweep” (a fast move that hunts stop-losses), followed by confirmation.
When it buys: after a sweep + break-of-structure + displacement confirmation across multiple timeframes.
How it sizes:risk-based sizing from stop distance:
risk a small % of equity (e.g. 0.5–1.0%), compute quantity from stop-loss distance, then cap notional.
How it takes profit: can emit an exit signal and/or target levels; the engine enforces stop-loss / take-profit rules.
On testnet we keep LR intentionally conservative: small risk per trade, hard notional caps, and cooldowns to avoid over-trading.
Profit & withdrawals
How profit is captured (and how scholarships are funded)
Unrealized PnL = profit/loss on positions that are still open (it moves with the market).
Realized PnL = profit/loss that has been locked-in because a position (or part of it) was closed.
Taking profit happens when the engine closes a position because:
a take-profit price is reached, or
a trailing stop locks in gains after price moves up, or
the strategy emits an explicit exit signal, or
risk rules trigger a safety exit (rare, “break glass”).
Where does the money go? Profit stays inside the trading account on the exchange.
The bot does not automatically withdraw to a bank account.
Scholarships are funded by an operator decision: you transfer a portion of realized profits using your normal finance process.
Sizing rules
Is position size fixed or dynamic?
Both exist—depending on the strategy:
Fixed-notional sizing (often Grid/DCA): “Spend $25 per entry” → quantity changes with price.
Risk-based sizing (often LR): “Risk 1% of equity to the stop-loss” → quantity depends on stop distance + equity.
Equity-fraction sizing (often Trend): “Use 2% of equity” → quantity scales with account size.
Always-on safety guard: regardless of strategy, the risk manager can block trades if they violate guardrails
(max risk per trade, max open positions, daily loss limit, drawdown limit, stop-loss requirement, etc.).
What is “ROI mode”?
ROI mode = “play to win the contest”
ROI mode is a more aggressive configuration profile. It doesn’t mean “ignore safety”—it means:
Stronger focus on higher upside moves.
More reliance on trailing stops (let winners run).
Different risk limits vs the conservative default.
You’ll see a badge on the dashboards: Mode: ROI or Mode: BASE.