How Simfolio analyzes portfolios
The data, investment rules and assumptions behind historical analysis and simulated outcomes. Explore the overview, or search for a tool, metric or ticker.
Search the overview, technical methods and source catalogues.
Analysis uses the dates covered by every required asset, currency and reference series. Constructed histories extend selected exposures with earlier observed data or modeled returns. These extensions describe an exposure before a fund existed, not returns earned by that fund. Source catalogues record the underlying histories and assumptions.
Results reflect both asset allocation and the rules used to maintain it: rebalancing, contributions, withdrawals and applicable costs. Strategy Composer keeps investment rules active through the analysis. Reporting currency and inflation settings determine how values are expressed; individual tools apply the assumptions described in their methods.
Backtesting estimates how a portfolio would have performed under the selected investment rules. Wealth includes contributions and withdrawals; investment-return measures separate those flows. Sequence of Returns Risk examines the effects of return order. Fixed-horizon boxplots in Rolling Metrics compare compound annual growth rates (CAGR) over different window lengths. Historical results remain specific to the available sample.
Simulation explores possible outcomes under explicit return assumptions. The Default Model resamples historical days, the Frontier Model adapts a combination of return models to your portfolio’s history, and Custom uses your selected specification. Percentiles and funding probabilities describe conditional outcomes. A percentile line can combine different scenarios over time; it is not one continuous investment path or a guaranteed range.
Specialist tools examine return drivers, allocation choices, spending and cash-flow valuation. Their methods distinguish historical comparisons, simulated estimates and constant-rate calculations. An optimized allocation or withdrawal rate depends on the selected data, constraints and assumptions. Loading a saved portfolio does not automatically transfer every Research setting.
Saved portfolios retain analysis inputs. Portfolio Reports, Simulation Reports and compatible AI applications provide ways to use and review the same underlying analysis. Their results retain its assumptions; AI-generated explanations should be checked against the inputs and returned results.
Data revisions, incomplete histories and modeled extensions affect historical evidence. Future events can differ from the conditions represented in a model, and actual costs and taxes can differ from estimates. Read each result with its relevant assumptions. Simfolio provides research and educational analysis; investment decisions remain the investor’s responsibility.
Each analysis aligns market, currency, and economic histories; their overlap limits the usable sample and requested dates, including required comparison series.
A market-data provider supplies prices, distributions, splits, indices, futures, and cryptocurrency histories. Adjusted histories incorporate relevant corporate actions and distributions.
FRED supplies Treasury yields, short-term rates, inflation, and foreign exchange, including OECD, Eurostat, and IMF series. Historical currencies also use Federal Reserve, Federico–Tena, and Bank for International Settlements data; commodities use the World Bank Pink Sheet.
The Kenneth French Data Library supplies academic factors and research portfolios; Global-q supplies the Hou–Xue–Zhang specification. Custom factors use selected ticker-return histories.
This product uses the FRED® API but is not endorsed or certified by the Federal Reserve Bank of St. Louis.
Early foreign-exchange histories interpolate lower-frequency observations and may convert predecessor currencies. Supported regional histories may use reference-series proxies before local coverage. Reconstructed histories can use data or revisions that were unavailable on the historical date.
Constructed histories extend exposures with earlier funds, indices, research portfolios or modeled returns. Backfill extends the history using earlier observed data or modeled returns; it does not represent returns earned by a fund before its inception.
The source catalogues identify coverage dates, splices, exposure adjustments, fees and construction assumptions. Observed instrument history, observed substitutes and modeled extensions should be interpreted separately.
Explore constructed history sources
Uploaded daily returns become a reusable ticker. Optional backfill and forward fill extend coverage with a selected reference history, combining uploaded data with assumed extensions.
Target weights define allocation; holdings drift with returns between trades. Calendar schedules and drift bands trigger rebalancing to target.
Absolute bands measure percentage-point drift; relative bands measure drift as a share of target weight. For a 20% target, a 5-percentage-point band and a 25% relative band both place the upper threshold at 25%. Separate upper and lower thresholds allow asymmetric rules.
Asset-level leverage, financing and fees modify return histories. Annual Drag adds a recurring portfolio-level cost.
One-time or recurring contributions and withdrawals use currency amounts or percentages of portfolio value. Inflation indexing preserves the purchasing power of fixed amounts. Flows are applied after the day’s investment return; a fixed contribution on a payment date is applied before a dynamic withdrawal.
Dynamic withdrawals recalculate spending from current wealth, the remaining payment schedule, a terminal-value target, and an assumed return, with optional minimum and maximum payments. Guyton–Klinger applies inflation adjustments and guardrails; time- and wealth-based policies apply selected spending rates to successive periods or wealth bands.
Results distinguish requested withdrawals, funded payments and shortfalls. A requested payment that exceeds available funding is not treated as fully paid.
Returns are converted to the reporting currency before aggregation. Inflation-adjusted wealth and returns express purchasing power, separately from cash-flow indexing.
Time-weighted returns exclude external cash-flow timing; money-weighted returns reflect it. Risk-adjusted measures use the reporting currency’s configured risk-free reference series.
Each portfolio has its own all-in one-way trading-cost assumption, defaulting to 0%. The configured percentage covers spread, slippage and commissions and is multiplied by one-way turnover (half the sum of absolute weight changes). For example, 0.15% means 15 basis points per unit of one-way turnover. New runs do not add an automatic spread estimate or fallback cost. Previously saved results retain the assumptions used when they were calculated.
These are modeled execution costs, not actual fills; they exclude order-size-dependent market impact.
Tax estimates show how realized gains affect performance under the entered tax rates. Changes in estimated liability adjust reported returns; original dollar balances and investor cash flows remain unchanged. After-tax ending value is reported separately, rather than treating estimated tax as a cash payment from the portfolio.
The calculation tracks cost basis in twelve monthly short-term pools and one long-term pool. Realized gains are estimated proportionally from the portfolio’s rebalancing turnover and gross executed cash flows. Tax rates may be entered as one effective rate or separate short- and long-term rates. Annual losses are netted and carried forward.
These estimates use supported cost-basis data, not a complete jurisdiction-specific tax system. Efficient Frontier and Simulated Efficient Frontier exclude taxes. Other standalone tools are pre-tax unless their method explicitly includes tax.
Wealth includes investment returns and external cash flows. The definitions below distinguish performance, risk and the effect of cash-flow timing. Calendar and rolling measures apply these ideas to different parts of the selected history.
Daily historical volatility uses sample standard deviation and annualizes with 252 trading days. Risk-adjusted measures use the reporting currency’s configured risk-free reference; downside measures use the stated target. Simulated allocation comparisons use their fitted return distributions rather than a historical portfolio replay.
In Rolling Metrics, Summary, Distribution and Scatter use the selected rolling window. Boxplot instead compares CAGR over complete fixed horizons of 1, 2, 3, 5, 7, 10, 15, 20, 25 and 30 years, plus 50 and 100 in Simulation when the simulated period is long enough. A year is 252 daily returns. Each box spans the 25th to 75th percentile with a median line and minimum-to-maximum whiskers. Historical observations are rolling windows on the portfolio path. Simulations pool all finite path/window outcomes, not one terminal value per path.
Overlapping windows share observations. Their results are not independent evidence or guaranteed future ranges. Read tax liabilities separately from cash paid, and requested withdrawals separately from funded payments.
Strategies combine asset and group blocks with equal, specified or inverse-volatility weights, conditional branches and ranked filters. Inverse-volatility allocations and ranked filters operate on individual assets. Signals use price and return information, including moving averages, volatility, drawdown and relative strength, rather than fundamental company data.
Scheduled trading reviews rule-selected allocations at the chosen frequency. Threshold trading checks absolute allocation drift daily and trades when it exceeds the configured threshold. A signal observed at a close can affect only the next return interval. When indicators lack enough observations, the selected warm-up policy either holds zero-return cash or requires complete history before the analysis can proceed.
During simulation, each path maintains its own indicator and holding state. Rules respond to that path’s simulated prices, and inverse-volatility weights continue to change with those observations. Strategy rules therefore remain active rather than being replaced with a single historical strategy-return series or a frozen allocation.
Strategies are retained as saved portfolio inputs and can be loaded into research. Uploaded custom return-series tickers remain a separate way to supply a return history.
Compare original history with reordered return blocks under the selected contribution schedule. Optional CAGR variation changes overall growth, so differences reflect sequence and any selected growth variation. Reordered paths are hypothetical.
Default draws historical trading days independently and applies each draw to all assets together. This preserves observed same-day relationships between assets but removes the original time ordering.
Before sampling, it shrinks each asset’s estimated mean log return toward zero according to its uncertainty. The adjusted historical daily distribution then provides the model for future returns.
Frontier is Simfolio’s adaptive simulation model. It fits a combination of return models to your portfolio’s history, learning how much weight to give each approach from its ability to describe outcomes in earlier historical tests. The mix adapts to the supplied history instead of relying on one fixed description of returns.
This learned combination is what distinguishes Frontier from the Default Model’s historical resampling. Each scenario follows one fitted specification over its full horizon, and the resulting scenarios form the combined distribution of possible portfolio outcomes.
The published historical evaluation found lower aggregate distributional error than its benchmark under the study’s tested conditions. That result concerns the accuracy of simulated distributions; it does not predict higher investment returns or guarantee future accuracy.
Percentiles describe the spread of simulated wealth, returns, and other outcomes. Plotted percentile lines summarize values separately at each date and can combine different scenarios over time; a band does not imply that a scenario remains inside it throughout the horizon. Funding probabilities measure how often scenarios meet the selected requirement. These are conditional model outputs, not guaranteed future ranges.
Future inflation follows one deterministic path from the latest annual rate toward its long-run historical rate. That path supports purchasing-power reporting, indexed cash flows, and real risk-free returns; scenarios do not simulate inflation independently.
Custom simulation selects a model specification from supported combinations. The available structures are Historical Resampling, Filtered Returns and Regime-Switching Normal. Compatible choices determine the mean, volatility, innovation, resampling and dependence assumptions of the return process.
A recipe’s availability indicates that the implementation supports that combination. It is not evidence that every recipe is more accurate than Default or Frontier. The published Frontier study does not validate arbitrary Custom selections.
Simulation Validation fits a selected model using one part of the available history, then compares its simulated outcomes with what happened in a later, held-out period. You can choose the split dates and test Default, Frontier or a supported Custom specification.
The distributional error score compares the predicted range of returns with the realized outcome; a lower score indicates a closer fit on the same dates and assumptions. Backtest percentiles locate realized metrics within simulated distributions and are not accuracy percentages.
One user-selected historical test does not establish general model superiority. This tool is separate from the broader published Frontier evaluation.
Loading a saved allocation does not import every Research workspace setting. Each tool uses the inputs and assumptions shown in its own method.
| Analysis | Portfolio assumptions | Costs and reporting |
|---|---|---|
| Backtesting and Simulation | Selected holdings, rebalancing rules and cash-flow policies. | Apply the selected reporting, cost and tax settings. |
| Efficient Frontier and Simulated Efficient Frontier | Constant daily weights on aligned historical returns or completed simulation assumptions. | Trading costs, cash flows, taxes and Annual Drag are excluded. |
| Factor Regression | Relates portfolio returns to the selected factors. | Nominal, pre-tax returns. |
| Other standalone tools | Use their own inputs and methods; workspace settings are not automatically inherited. | Pre-tax unless the method explicitly includes tax. |
| Time Value of Money and NPV & IRR | Specified cash flows, timing and rates. | Constant-rate calculations, without simulated market uncertainty. |
Regression relates nominal, pre-tax returns to academic or custom factors. Best Fit compares supported specifications on common observations using adjusted R². Loadings estimate exposures; alpha and fit statistics show what the model explains and leaves unexplained.
Academic specifications include CAPM, Fama–French, Carhart, and Hou–Xue–Zhang, with BAB and QMJ alternatives. FF8 and FF10 are exploratory combinations of published factors, not established named models.
PCA decomposes the selected correlation or covariance matrix into mutually uncorrelated components. Loadings show asset contributions; explained variance measures each component’s share of variation. Components are statistical patterns, not predefined economic factors.
Compares daily constant-weight allocations before trading costs, cash flows, taxes and portfolio-level drag.
Using the currently displayed completed backtest, the optimizer searches for an allocation that maximizes compound annual growth rate (CAGR), minimizes annualized volatility, or maximizes the Sharpe ratio. The Sharpe ratio measures excess return relative to volatility.
The optimizer uses the complete aligned history to estimate performance at constant daily weights.
Results compare the selected allocation with the original on all three measures and report the objective value, solver status, and constraint checks. The heuristic search does not establish a global optimum or predict future performance.
Reuses the completed simulation’s joint return and covariance assumptions to estimate outcomes at constant daily weights. Trading costs, cash flows, taxes and portfolio-level drag are excluded.
The selected objective is to maximize CAGR, minimize volatility, or maximize the Sharpe ratio at the configured percentile.
Results compare the optimized allocation with the current allocation under the fitted simulation assumptions. The finite allocation search is not a proof of a global optimum.
The standalone tool reports historical results only.
At each optimization date, prior-dated returns, the selected model, lookback, and bounds determine target weights for subsequent returns, separating estimation from performance. Research simulations using historical optimization start from the current underlying holdings and apply the selected rebalancing policy without refitting the allocation within future scenarios.
The objectives below distinguish the risk or return measure each method seeks to improve.
An objective value describes the fitted optimization problem. Estimation choices, constraints and the selected history affect the result; a successful solver does not establish a universally best allocation or future performance.
Solve for a starting balance, regular payment, rate, term or ending balance. In the equation, PV is the starting value, P the regular payment, r the rate per payment period and n the number of periods. The timing factor t is 1 for end-of-period payments and 1 + r for beginning-of-period payments. At a zero rate, FV = PV + P × n.
Effective annual rates convert to the selected payment frequency; nominal rates also use an explicit compounding frequency. Loan repayments and extras reduce the balance, with a capped final payment and an explicit remaining balloon.
Fractional solved terms and detected rate solutions are labeled. Rates and cash flows are assumed as entered; the calculation does not model market uncertainty or provide statutory APR disclosures.
Enter outflows as negative and inflows as positive amounts. The initial flow is the undiscounted time-zero amount and may be zero. Each following row occurs at its numbered period; zero rows preserve skipped periods.
In the NPV equation, Cₖ is the cash flow in period k, d the effective annual discount rate, f the number of periods per year and n the final period. IRR solves for the rate at which NPV is zero; a periodic IRR annualizes as (1 + periodic IRR) raised to f, minus 1.
All validated roots detected within the displayed search range are reported, with non-uniqueness and numerical limits explicit. Recovery is the first actual period when cumulative cash flow becomes nonnegative after a negative initial flow. Later reversals are flagged; payment timing is not interpolated.
Log-growth estimates use asset CAGR and the cash rate, subtract a quadratic volatility penalty, and charge borrowing spread and incremental expenses on exposure above one times capital. The optimum maximizes modeled growth under these assumptions.
The withdrawal-rate curves use a constant annual spending amount as a share of starting wealth, paid in equal daily installments after investment returns. With inflation adjustment, this amount is constant in purchasing power. A safe withdrawal rate funds the payments through the chosen horizon; a principal-preserving rate also retains the starting wealth at the end, on the selected real or nominal basis.
The success threshold is the share of evaluated starting windows that meet the requirement. It is a historical frequency under the selected assumptions, not a statistical confidence interval or a guarantee of future success. Overlapping windows share observations.
Historical-only results use complete windows. Long-term continuation or a custom CAGR can extend incomplete windows with assumed growth. A selected custom withdrawal policy is evaluated under its own payment rules; it is distinct from the constant-spending curves.
Saved portfolios retain their inputs and can be loaded directly into an analysis, including Strategy Composer rules.
Portfolio Reports review the selected portfolio and benchmark at the configured cadence, including event reports. Each report retains the data and portfolio rules of its underlying analysis.
Simulation Reports schedule new analyses of the selected portfolio. Each report retains the data, portfolio rules and model assumptions of its underlying analysis.
The Model Context Protocol (MCP) connection lets compatible AI applications use supported Simfolio features within the account’s permissions. The underlying methods still apply; check generated explanations against the inputs and returned results.
Historical results describe the selected sample and remain exposed to data revisions, survivorship, incomplete coverage, and proxy-construction risk.
Structural changes or events absent from the fitted history can fall outside simulated ranges. Optimization depends on the sample, objective, constraints, and estimation method; shrinkage and separating estimation from evaluation address specific errors without eliminating uncertainty.
Simfolio provides research and educational analysis, not investment advice.
Modeled spreads are not actual fills and exclude order-size-dependent market impact. Tax reports are estimates, not tax returns. Investment decisions remain the investor’s responsibility.
Enter a ticker like TMFSIM to use its prebuilt history. Before launch, Simfolio applies the fund’s stated daily leverage to its benchmark or a matching proxy. Observed fund prices or note values take over from launch.
For fitted funds, Simfolio uses dated operating expenses from SEC filings and adds back fees embedded in a fund used as the benchmark proxy. It estimates the financing spread and multiplier from observed returns, bounding the multiplier between 1.00× and 1.50×. It ranks the fits by compounded log-return error across every pair of trading dates within later validation windows. A final window checks tracking and ending wealth. The prelaunch history uses the launch fee or the earliest verified fee. Fits and filings are reviewed monthly.
GLLDSIM and its alias SHNYSIM use BMO’s published note terms for the prelaunch history.
Enter QQQ?L=3 for 3× daily exposure to QQQ. The financing multiplier and spread are averages across fitted funds, calculated separately for long and inverse leverage. The operating expense is the median of benchmark-family fee medians for each direction. The fee scales linearly from zero at 1× to its cap at 2× for long leverage, and from zero at 0× to its cap at −1× for inverse leverage.
Add any override with &, as in QQQ?L=3&SP=0.50&E=1. Enter SP, E and UE in annual percentage points. The formula uses decimals, so E=1 enters the calculation as 0.01 per year.
L=3FR=CASHXSW=1.20SP=0.50E=1UE=0.20Trace the sources behind a constructed exposure history, including coverage dates, earlier substitutes, exposure adjustments and fees.
Source series and research-portfolio labels retain their published identifiers. In factor histories, RF is the risk-free return and Mkt-RF is the market return above that rate; adding them gives the total market return. Daily exposure multiples describe the modeled return exposure, while financing spreads and fees are annual rates unless stated otherwise.
103 of 103 histories · 1–12
Find the purpose, coverage, sources and accrual assumptions of cash and interest-rate reference histories. These reference series are not investable fund returns.
Entries use either published index levels or interest rates converted into a return index. Modeled accrual is the calculation of interest earned over time from those rates. Coverage dates describe the return history; any extension beyond the last observed rate follows the stated accrual rule. Source identifiers and links are retained for verification.
19 of 19 histories · 1–12
Every required asset, currency and reference series must overlap. A portfolio with a shorter history can therefore have a later usable start date.
Each tool uses the settings shown in its own inputs. Loading a saved allocation does not automatically import every Research setting; cost, tax and cash-flow assumptions can differ by tool.
Wealth includes contributions and withdrawals as well as investment returns. Time-weighted return separates external cash flows, while money-weighted return reflects their amounts and timing.
Inflation adjustment expresses wealth and returns in purchasing-power terms. It is separate from inflation indexing of contributions and withdrawals.
A percentile line summarizes outcomes separately at each date and can combine different scenarios over time. It is not one continuous path, and a scenario need not stay inside a displayed band.
Success means funding the required payments through the chosen horizon. Principal preservation also requires retaining starting wealth at the end. A historical success frequency is not a guarantee of future success.
Estimated tax liability adjusts reported after-tax performance. It is not treated as a cash payment from the portfolio; original dollar balances and investor cash flows remain unchanged.
The source catalogues list coverage dates, earlier substitutes, construction assumptions, fees and source links. They distinguish observed history from modeled extensions.