How to use NHLFutures

The date and source behind the page

The dashboard opens on a selected benchmark snapshot, with its publication date and status shown nearby. Team rates can come from a different source season than that snapshot, so it is worth checking both labels before reading too much into a move. Selecting an earlier date changes the benchmark view; it does not rewrite today’s team rates as if they had been known then.

Index points and percentages

FSPI values appear in their native index points, so 7,500 is displayed as 7,500. Futures quotes may use a different contract convention, while percentages on the site refer to the probability of a named event. When you compare the dashboard with an export or exchange quote, check that the units and end dates line up.

An example team row

Imagine a row labeled current index 7,500, month-end forecast 7,650, and forecast interval 7,350–7,950. The center is 150 points above current, but the interval includes a decline. That is a distribution of possible outcomes, not an assured 150-point gain. If the row instead shows 65%, first identify the named event: an event probability is not the same unit as the index.

The word “month” also needs a quick check: one view may mean the next 30 days while another refers to a named calendar month. Two monthly-looking numbers with different end dates are measuring different periods.

What each part of the site is for

• Dashboard: benchmark context, team selection, and indexed history.
• Team data: public team rates, with protected projection columns.
• Games: schedule and matchup context for the selected schedule date.
• Model Studio: model-wide assumptions and future or historical runs.
• Team Explorer: a selected team and its bounded scenario assumptions.
• Model Compare: compare two configurations over the same horizon.

Trying a scenario

Choose a team and inspect the reference. Change one supported assumption, run the scenario if your account is entitled, then compare the delta with the unchanged reference. Player controls are future-only. A reduced player-impact setting describes an assumption; it does not report an actual injury or trade.

Comparing a scenario with the reference

It helps to begin with a specific question, such as, “How much does this team’s outlook change if one player contributes less over the next few games?” Start from the reference forecast, choose the future horizon, and change that one assumption. Keeping the reference beside the scenario makes the difference easier to understand.

Inspect the scenario delta beside the reference, then check whether uncertainty or downstream probabilities changed as well. Save a supported configuration with a descriptive name, rather than a claim such as “winning model.” When comparing configurations, make sure the dates, mode, and profile are aligned. An older saved configuration may not contain a complete cached result; read the execution status shown by the tool.

Unavailable and locked results

An “illustrative preview” contains demonstration data. “Shadow validation” means the output is being monitored, while “production” means that a particular publication was approved for release. The timestamp still tells you whether that publication is current. Blank and unavailable results mean no valid value was returned; they do not represent a forecast of zero.

What the access levels include

Free access covers public benchmark and research information. Plus covers available published default-model projections. Pro adds custom simulations, saved configurations and historical evaluation; legacy Premium and admin access retain Pro capabilities. Billing activation and actual data availability are separate from those access definitions.

Access and data availability are separate. A paid account cannot fill in a missing source or turn an experimental model into a validated one. If a run fails, the page should continue to distinguish that failure from the last successful result; an empty result should not be read as a neutral forecast.

Keep a traceable comparison

Save supported configurations and compare like-for-like results. Watchlist preferences do not mean scheduled notifications are already being delivered. Research pages are public and can be read without signing in.