One strategy wants to sell ten shares. Another wants to buy ten. Averaging their equity curves makes the combination look less volatile, but the actual account might need no market trade—or two trades with separate fees. Which arrangement are you studying?
This is part of the research question. You need an explanation of possible complementary returns and an account model showing how one pool of capital funds conflicting targets. A smoother curve acquires a useful comparison only after those connections are specified.
Check whether both strategies qualify as inputs
You will review the individual evidence, freeze simple comparisons on equal capital, reconcile opposing targets, and save a T20 plan that permits a conclusion of no incremental value. Dynamic weighting and an optimizer are not the starting requirements.
Each strategy supplies rules and versions, data and permissions, signal-availability times, holdings, fills and cash, costs, benchmarks, historical and forward status, and failed attempts. Return to the individual study if these are incomplete. A second rising curve cannot repair an unexplained first backtest.
C2-A and C2-B could be candidates, but their complete real single-strategy artifacts do not yet support this comparison. We therefore use an independent synthetic mechanism package, without claiming verified C2 long-term performance, diversification, or operating readiness. A factor strategy likewise needs its own evidence before entering the combination.
Distinguish signals, strategies, and the account
Combining signal columns into one score and position rule is signal combination. Allocating capital between separately specified entry, exit, and cost rules is strategy combination. The account layer determines how their targets become actual orders. A result at either earlier layer does not replace the account ledger.
QuantConnect's framework separates signals, target construction, risk management, and execution, while describing limits to the strategy styles its defaults fit.[1] That is an architectural reference, not evidence that CoTrading or your tool supports the same process.
Write the proposed benefit: different return mechanisms or holding periods, offset cash demands, or fewer unnecessary trades in one account. Then name competing explanations and failure modes: common holdings, industry or style exposure, simultaneous selling under stress, and shared data failures. Different names do not establish different risk sources. Low historical correlation does not prove independence.
Allocate one budget before constructing targets
C10-A-v1 uses synthetic dollars, long-only positions, and no leverage. Signals arrive before their corresponding W1, W2, and W3 execution windows, with supplied prices of $10, $12, and $11. Process them only at the next specified window, without backdating fills to signal prices. Inputs and expected records are in the editable package. Do not join A to the later B account.
The actual account begins with $1,000 cash and 40 X shares at $10, totaling $1,400. Virtual strategy ledger T owns $400 cash and 30 shares; R owns $600 and ten shares. Each initially has $700 equity. Equal capital budgets do not mean equal stock ownership, and they do not reset to $700 at every window.
This case permits netting in one account. Opposing intentions first transfer virtual shares and matching virtual cash at the window price. Only the remainder trades externally. Internal transfers have no market fill or commission. Each external order costs $1 on its first actual fill; later fills of that same order do not repeat it. There are no distributions, taxes, or other costs. Shares are integers and neither strategy can borrow the other's cash.
At W1, T's target falls from 30 to 20 shares while R's rises from ten to 20. Transfer ten shares from T to R and $100 of virtual cash from R to T. Each ledger now has $500 cash and 20 shares. The actual account retains $1,000 cash and 40 shares; no exchange confirmation for ten shares exists.
With two separate real accounts, that free internal offset cannot be assumed. Under this case's fee schedule, T sells ten shares and R buys ten, each paying $1. Combined equity is $1,398 rather than $1,400. The $2 difference comes from execution policy, not stock-picking alpha. Actual market conditions would need separate fill evidence.
Netting does not remove execution uncertainty
At W2's $12 price, T targets zero and R targets 50. Transfer T's remaining 20 shares internally first. The actual account then needs to buy only ten additional shares. The exercise permits just four to fill, leaving an actual position of 44 shares and six still open.
Actual cash is $951. T owns $740 of it and R only $211. R's remaining order reserves $72, leaving $139 available for another commitment. The account's larger cash total does not authorize R to use T's budget. At W3 the prescribed process confirms cancellation, then sells the existing 44 shares, ending at $1,434 cash. G92 opens the full event and contribution record.
T20 must therefore specify netting, partial fills, open-order treatment, and where a paused strategy's budget goes. The package also supplies separate insufficient-cash and uncanceled-old-order rejection branches. Two equity curves without strategy ownership, open orders, and reservations cannot establish the shared-account model.
Only after specifying implementation should you compare performance. Otherwise the attractive curve may describe an arrangement that cannot be executed under the supposedly unchanged rules.
Compare returns on the same total capital
The independent C10-B mechanism starts from $1,000. Its complete six-month synthetic net returns appear below; underlying synthetic purchase and sale prices, actual commissions, and transactions remain in the package. T and R are not validated economic strategies. All six months are already known. The first three, fourth, and final two play development, validation, and historical teaching-holdout roles, not unknown future evidence.
| Month | T net | R net | Equal capital budgets | Candidate: 25% T / 75% R |
|---|---|---|---|---|
| M1 | 4% | 1% | 2.5% | 1.75% |
| M2 | −4% | 1% | −1.5% | −0.25% |
| M3 | 4% | 1% | 2.5% | 1.75% |
| M4 | −4% | 1% | −1.5% | −0.25% |
| M5 | 4% | −2% | 1% | −0.5% |
| M6 | −4% | −2% | −3% | −2.5% |
At each month's start, divide current equity into cash budgets. Each strategy buys its own synthetic asset and sells fully back to cash at month-end, then budgets are reassigned. Fractional shares are allowed; cash earns zero. Each purchase and sale charges five basis points of actual value, with no additional costs. Reallocating cash does not itself invent another securities trade. Strategies that retain holdings across months cannot inherit this assumption of no additional reallocation trading.
The four comparisons are T alone, R alone, equal budgets, and the sole candidate, 25/75. Each has only $1,000 total capital. The candidate reduces allocation to the more variable T; it does not claim 75% is optimal. Register these two combinations without searching further weights after seeing results. Equal budgets are not equal risk, and overlap and liquidity still require review.
In the equal-budget first month, T's $500 becomes $520 and R's becomes $505, totaling $1,025. Split that total into $512.50 each for month two. Their endpoints are $492 and $517.625, totaling $1,009.625. Net-return weighting works here because of the specific budgets, cost model, and monthly cash reset. It cannot replace the fixed-fee netted execution ledger in A.
A steadier candidate can still fail its criterion
| Same-capital alternative | Six-month ending dollars | Month-end maximum drawdown | Annualized sample volatility | Mechanically annualized CAGR | Sharpe, monthly risk-free reference 0 |
|---|---|---|---|---|---|
| T alone | 995.207676 | 4.306954% | 15.178933% | −0.956168% | 0 |
| R alone | 999.396091 | 3.960000% | 5.366563% | −0.120745% | 0 |
| Equal budgets | 998.649985 | 3.499550% | 8.049845% | −0.269821% | 0 |
| Candidate 25/75 | 999.360871 | 3.230031% | 5.531727% | −0.127785% | 0 |
Recalculate metrics from each complete monthly account path. Volatility uses the monthly sample standard deviation times the square root of twelve. Sharpe uses mean monthly differential return divided by sample standard deviation, then the same multiplier—not CAGR or averaged single-strategy Sharpes. All four arithmetic monthly means happen to be zero, so each Sharpe is zero despite different compounded results and drawdowns. Six-month annualization illustrates a calculation, not a long-term forecast.
The frozen mechanism criterion requires the candidate to have lower drawdown than equal budgets and no worse cumulative performance than either individual reference. It has lower observed drawdown but ends slightly below R alone. Retain the conclusion that the additional combination has not passed this criterion. Deleting R afterward would change the test to favor the candidate.
CFA Institute distinguishes measurement, attribution, and appraisal and emphasizes their data and inference limits.[2] The table measures results. Explaining improvement and whether it compensates for added management complexity requires economic and implementation evidence beyond it.
Reprice costs and common shocks from the original inputs
Equal budgets incur about $6.117696 in base-case fees and $12,235.391034 in two-sided trade value. Relative to the initial $1,000, cumulative turnover is about 1,223.5391%. This does not mean twelvefold turnover every month. Monthly turnover uses that month's starting equity; the cumulative figure uses initial capital. Fees already reduce net returns and must not be deducted again from $998.649985.
The package freezes three additional scenarios: raise fees to 15 basis points at unchanged raw prices; retain base fees but reduce every actual exit sale price by 1%; or retain base fees and reduce only the final month's sale prices for both assets by 10%. Recalculate fees, equity, and metrics for each, rather than subtracting one generic percentage from the old net curve.
The candidate's endpoints are approximately $987.440194, $940.878422, and $899.424784 respectively. Under higher fees, its recalculated Sharpe is about negative 0.4343. Weighting the standalone Sharpes 25/75 gives about negative 0.3753 instead. They differ because the combined risk denominator must come from the combined sequence.
The six-period sample correlation between T and R is exactly zero, yet both lose in M6 and both are exposed to the joint stress. That neither establishes independence nor estimates a stable correlation. Common stocks, styles, data, or order channels also belong in T20; one correlation coefficient cannot stand in for all those checks.
Change a price and distinguish the ledgers
First explain why W1's internal ten-share transfer is not two real filled orders. Then start an independent W2 variant from the post-W1 ledgers: each has $500 cash and 20 shares. T targets zero and R 50, but the W2 price is $13. Transfer 20 internally, then permit only four external shares to fill, with the first-fill $1 fee and six remaining active. Find each strategy's cash, actual account equity, and R's available cash after reservation.
Reference answers · reveal
Answer guide: T receives $260, ending with $760 cash. R pays $260, then $52 plus $1 externally, ending with $187 cash and 44 shares. Actual account cash is $947; 44 shares at $13 give equity $1,519. R's equity is $759. Reserve $78 for its six open shares, leaving $109 available. T's $760 plus R's $759 reconcile to $1,519. The 50-share target remains unfilled. Do not reuse the base case's $951 cash or $72 reservation.
For the explanation, W1 changes strategy ownership, while actual account shares and cash remain unchanged. It creates no exchange fill or new value. The separate-account comparison requires real buys and sells under its given assumptions and consequently pays $2 more.
Save T20 before account acceptance
Retain both single-strategy qualification records, proposed return mechanisms and alternatives, common exposures and data dependencies, one capital budget, matching dates, currency, valuation, costs, equal-budget baseline, and one candidate. Specify allocation frequency, cash constraints, paused-budget treatment, viewed partitions, every attempt, metric definitions, rejection criteria, and small-sample account expectations.
Check that every budget draws from the same capital once; targets remain separate from fills; internal transfers do not pretend to be market trades; costs enter once; and missing real evidence remains missing. Keeping only a simple baseline, continuing simulation, or declining to combine are valid outcomes. A multi-strategy deliverable does not require acceptance of another strategy.
Make the comparison fair before claiming an increment
Review each input study, freeze a fair comparison, and make targets, execution, and cash explain the combined results. Complementarity is a hypothesis, and lower volatility is only one feature. Additional complexity needs evidence of a worthwhile increment under the same capital, definitions, and constraints.
References
- QuantConnect: Algorithm Framework Overview, Important Terminology, Strategy Styles, and Separation of Concerns. Architecture reference, not a CoTrading feature or optimal-allocation claim.
- CFA Institute: Portfolio Performance Evaluation, 2026 public overview. Measurement, attribution, appraisal, and their evidential limits; not the paid full reading.
Sources inspected October 30, 2026. C10-A and B are separate synthetic exercises. AI-assisted draft; independent AI full-text and bilingual review is complete. Human financial review and target-reader testing have not been performed. Real C2 single-strategy artifacts, real multi-strategy evidence, and RUN acceptance remain incomplete.