Skip to main content
A Tilt is a rules-based portfolio: a set of construction rules and the constituents they produce. Tilt applies the rules to a universe of securities, ranks the securities that pass, sizes the positions, and returns the constituents. Each rebalance runs the same rules on fresh data. You can write the rules yourself or describe the Tilt to Tilt Agent, which writes them for you. You can read and edit every rule either way. Marketplace publishing creates a public version of a Tilt. Submit as an Index starts a separate index-submission review. Read working copies and published versions for version labels, Marketplace listing, and sleeve use.

The construction pipeline

The construction pipeline: universe, derived fields, selection rules, position sizing, ticker overrides, rebalance and concentration, constituents
Each stage is a section on a Tilt’s Methodology tab: The result is the Tilt’s constituents: the selected securities and their weights on the evaluation date.

Find your way around

Select a Tilt in Library to open its Tilt view. In chat, select View details on a Tilt card or select a Tilt reference in a sent message to open the same view without losing your place in the chat. Use Chats to find or switch conversations. Keep the Tilt open and check Looking at above the message input before asking about its rules or results. On your phone, chat opens over the Tilt; close chat to return to it.
The Tilt view with numbered callouts: 1 close, 2 the Methodology and Portfolio tabs, 3 the backtest, 4 the methodology sections, 5 the actions menu, 6 the History panel

The Tilt view, with History open.

  1. Close returns to the chat. A Tilt opened from Library shows Back to Library instead.
  2. Methodology shows the rules and a backtest. Portfolio lists the constituents and weights.
  3. The backtest compares the Tilt with a benchmark over a period you choose, with risk statistics.
  4. The methodology sections hold the construction rules, one section per stage. Open a section and edit it. Each edit joins the change queue at the bottom of the Tilt view. Open the queue and select Apply to save the changes and re-evaluate the Tilt.
  5. The actions menu has History, Export to Excel, and Publish to Marketplace.
  6. History lists every methodology change, with who made it and when.
After opening a Tilt from Library, select Explore in chat and choose a suggested question or Ask anything. New Chat opens with an editable draft and the Tilt attached. Nothing is sent until you select Send. When you ask Tilt Agent for a change in the chat, it edits the same rules you see in the Tilt view.

Read the backtest

The backtest on a Tilt’s Overview tab runs the Tilt’s current rules over a period and compares the result with a benchmark. The benchmark button under the chart names the benchmark the backtest used. To compare with another fund, click it and pick SPY, QQQ or DIA, or search for any supported ETF by ticker or name, such as ACWI or EFA. Choosing a benchmark or a period runs the backtest again. A Tilt whose rules use a custom AI-scored column (enrichment()) can’t be run on past dates, so its chart is a backcast: it holds the Tilt’s latest saved basket, at its current weights, for the whole period. Read a backcast as a hindsight view of today’s basket. The rules didn’t pick those holdings on past dates. The summary beside the chart shows the Tilt’s Return, the benchmark’s return, and the Relative return between them, with these risk statistics: Historical tracking error describes this backtest only. It isn’t the tracking error of a live account or a sleeve, and it isn’t a target. It reads Unavailable when the period doesn’t have enough days with returns for both the Tilt and the benchmark. If the backtest leaves out tickers it can’t support, a note under the chart lists them, and the figures describe only the tickers it kept.

Turnover and trading costs

When a rebalancing equity Tilt reports turnover, it measures one-way turnover as half the total value bought and sold at each simulated rebalance, divided by the portfolio value immediately before that rebalance. It uses that day’s closing prices and the holdings after corporate-action adjustments. Price drift between rebalances counts: restoring the same target weights can require trades even when no ticker enters or leaves. Selling $10,000 and buying $10,000 in a $100,000 portfolio produces 10% one-way turnover. The two-way traded value is $20,000, or 20% of portfolio value. Turnover is Unavailable for buy-and-hold runs and for the long/short idea backtests labeled Shows the idea. Not investable as built. It is also unavailable when the simulated holdings before or after a rebalance lack valid same-day prices, or the pre-rebalance portfolio value is nonpositive. A supported run with only the initial purchase has zero subsequent turnover. The annual average also needs a positive observed duration. These are simulated trading measures; they differ from a live fund’s annual turnover convention. The equity backtest’s returns are before trading costs: For a separate cost estimate, specify a cost per dollar traded and apply it to both buys and sells. At an assumed 10 basis points (0.10%) per dollar traded, the $20,000 example costs $20, or 0.02% of the $100,000 portfolio. This is an illustrative assumption, not a Tilt default or a cost already deducted from the chart. Initial purchases are excluded from reported turnover, so include their cost separately when estimating the cost of starting a portfolio. When comparing monthly and quarterly rebalancing, use the same historical window, construction rules apart from the schedule, and cost assumptions. Compare both gross returns and estimated trading costs; rebalance frequency alone does not determine turnover. Advisor proposal turnover uses a separate two-way measure.

A first example

This Tilt takes US companies worth more than $5 billion with strong free-cash-flow margins, ranks them by revenue growth and return on equity, and holds the top 20 by free-float market cap: market_cap is in millions of US dollars.

Next steps

Expression reference

Fields, operators, and functions for each rule.

Selection and weighting

How filters, scores, limits, weights, and caps combine.

Build a thematic Tilt

Build and refine a nuclear energy Tilt with Tilt Agent.

Choose an integration

Compare Tilt’s interface, Tilt Agent over MCP, and the REST API design preview.