Getting started with asset weight configuration

Configuration is not a directive

PitchCopilot uses "asset weight allocation" to describe the model's relative confidence in different research directions. This word is often misunderstood, so it needs to be made clear first: it is not a real asset arrangement, nor a promise of income, but a set of virtual point research language. The platform treats each event as a research sample and uses a virtual fund pool to represent how the model allocates attention among core configurations, secondary configurations and observation items.

The advantage of this way of writing is that "I think" is changed into a comparable number. If a game has multiple possible directions, the model will not just give a single sentence, but will form different weights based on market implicit pricing, data premium baseline, actuarial premium rate, confidence interval and historical calibration performance. For relevant background, please read /guides/market-implied-pricing first.

Where does the weight come from?

The weights usually come from three layers of signals. The first layer is game data, such as team status, matchups, injuries, schedule density, and offensive and defensive quality. The second level is external consensus, which is the benchmark for market implicit pricing and goal ranges. The third layer is the calibrated model distribution, including wins, draws, losses, total goals and uncertainty. When the directions of the three layers are consistent, the allocation weight is usually more concentrated; when the signals pull each other, the weight will decrease or become the second choice.

PitchCopilot doesn’t package high weights into absolute answers. Human reviewers check whether the model relies too much on a single feature, ignores news context, and needs to be reminded of sample limitations in the report. After release, these configurations will be entered into /performance to review the quality using settlement rules and Brier scores.

Core, secondary options and observations

The core configuration represents the research direction where the model considers the signal to be the most complete. Secondary allocations are often supported by data, but the price gap, sample stability or risk control have not yet reached the main threshold. Observations are clues worth tracking but not suitable for virtual points research, such as a team's rotation risk, weather changes, or pregame lineup news.

Readers can think of the three as different levels of confidence rather than a ranking game. The core configuration may not hit every time, and the secondary configuration may also perform well; the key point is whether it meets the ex-ante definition in the long term. If you want to further understand the layering logic, you can read /guides/core-vs-secondary-config; if you want to know the settlement method, you can read /guides/settlement-hit-definitions.

FPL and Creator Apps

FPL players can use weight allocation to examine captain, transfer and defensive rotation risk, but it should not be relied upon as the only basis. If the model increases the weight of a team's offensive end, it also needs to be matched with player playing time, set-piece roles and schedule continuity. Creators can use weight changes to create pre-game narratives, such as "The model is more biased towards low-tempo scenes, but manual review retains the uncertainty of the lineup."

That’s why PitchCopilot emphasizes public performance. Without long-term records, configuration weights are just beautiful tables; with dashboards, readers can see the actual performance of different versions, different leagues, and different confidence levels.

The research significance of risk diversification

Another use of asset weight allocation is to show the degree of conservatism of the model. If only one direction of an event seems to be prominent, but other signals are weak, the model may still only give a medium weight; if multiple data layers support it at the same time, the weight will be increased. This dispersion is not to beautify the results, but to avoid hiding uncertainty in a strong judgment.

PitchCopilot's virtual point research will be publicly inspected after the game, so the weight design must be able to be verified over the long term. If the model frequently places high weights in low calibration intervals, Brier scores and virtual fund pool fluctuations will reflect the problem. If the secondary configuration is more stable than the core configuration in the long term, it means that the release threshold or feature weights may need to be adjusted. This feedback feeds into model version iterations rather than being ignored in a single report.

Readers can practically think of weights as a ranking tool rather than a list of actions. Look at the high-weighted direction first, and then check the reasons and limitations of the data; if the reasons cannot convince you, reduce its importance in personal research. This usage is particularly suitable for FPL players to manage their time and for creators to arrange the length of their articles.

Additional reading reminder

Configuration weights should also be read together with time points. The same event may have different information density at the beginning of the week and before the game, and manual review will be based on the time of release. If you review the report after the fact, please do not ask the model to know in advance the lineup information that appears later, but check whether the data known at the time is sufficient to support the weight.

FAQ

Are virtual points related to real money?

No. Virtual points are simply units of study used to compare different configuration assumptions.

Does a high weight necessarily mean it is correct?

no. The weights represent relative confidence and still need to be checked over time through public performance and settlement rules.

How do I get started?

Read /guides/start-here first, then use /explorer to search for event and team information.

This translation is for convenience only; the Traditional Chinese version prevails.