祝融说。

What Should Be Written for Requests Without Achievements

Gu Ning placed the original sixteen registered items back on the table. The eight achievement cards explain what was completed, but what should be written for requests that have no card? Tang Ke suggested filling in "incomplete"; Ye Cheng reminded her that not having found a card and not having been completed are still different, and that not completed does not mean the original window did not receive confirmation.

The previous chapter explained the entrance of selection; this chapter handles the outcomes that were left behind and not observed. An object may not have entered the achievement story yet still possess a known window outcome; it may also be known only that it was registered, with no knowledge of how it ended. Neither should be compressed into a single "did not succeed" column.

This chapter continues the thought experiment of the fictional Chengwan. R01–R08 already have five original-window confirmations and three that were not met; R09–R12 differ in stage or window; R13–R16 still lack original records. Because we have begun writing outcome states, we do not fabricate an eventual ending for the four missing objects.

R17 is still on Thursday, with the original deadline at 5 p.m. on Friday. It is not yet fully confirmed, so it cannot be marked as not held in the original window; insufficient material acquisition and observation not yet concluded are also different things. The state table must be able to preserve these distinctions if later material is to connect to judgment.

What Outcomes the Survivor Story Leaves Behind

To speak here of survivorship first means objects that ultimately remained in some preservation entrance and can continue to be told about; it does not require the request to have undergone physical life and death. The objects on the achievement cards passed the conditions of completion and display and can therefore be repeatedly returned to; the processes of those that did not enter may not be equally easy to access.

If one listens only to the surviving story, every item Gu Ning sees has an achievement. This may result from the preservation conditions rather than from every request originally having been completed. Chapter 13 already explained this entrance; this chapter asks what state the omitted objects are actually in and how they should be recorded so as not to erase them.

Requests that ultimately did not continue may have received original-window confirmation; requests that ultimately were completed may have passed through late confirmation or post-modification cooperation. The final state and the original event are two outcomes; neither lets the last part of the story—smooth or not—adjudicate the original window.

One can therefore preserve the fact of completion without letting the completion column govern all times and objects. The survivor story itself may be true; the question is whether it permanently removes from observation the object that initially entered the problem and its unresolved state.

The Window Has Not Yet Closed—Negation Cannot Be Generated in Advance

R17 currently has limited list-based replies, without a complete acceptance of all conditions in the first version. Gu Ning may record that confirmation has not yet been obtained to date, but what this book predicts is whether confirmation arrives by 5 p.m. on Friday; the current state does not rule out a later within-window acquisition.

If one fills it in as negative in advance, then changes it to positive after Friday's confirmation, there is a surface update that in fact treats an unconcluded observation as a known result. The updating of a probabilistic judgment and the continuation of outcome observation are different; this distinction cannot be eliminated by switching a binary column back and forth.

Consider a comparison: the other party explicitly states for now that it does not accept a certain period, but that does not necessarily mean the entire window can never yield first-version complete acceptance. Whether an irreversible exclusion exists requires additional conditions; one cannot adjudicate in advance merely from a tone of voice or a feeling of disappointment.

Preserving "not yet at deadline" does not mean refusing all early positive adjudications. If complete confirmation has been received within the window and the material can be verified, the receipt event can be known to have held; later changes in arrangement do not erase the original receipt. When receipt has not occurred, there is no symmetric basis for prematurely obtaining a negation of the entire remaining window.

After the Deadline Passes, Sufficient Outcome Material Is Still Required

When the window of a historical request closes, the time boundary for future opportunity has been reached; this does not guarantee that Gu Ning has acquired all necessary material within that boundary. When original communications are missing, she may still be unable to judge whether a specified confirmation was ever received.

R13–R16 maintain exactly this kind of material gap, but this chapter does not assign them specific messages or window processes. Knowing that a registration once existed is not enough to adjudicate the outcome under the current first-version definition; one must not assume that negation has already become true simply because a long time has passed.

By contrast, R01–R08 already have sufficient corresponding material and adjudications: five positive and three not held, with no need to reclassify them as unknown. Rigor does not mean revoking all old outcomes; it means letting the adjudicated and the not yet adjudicated each rest on their own grounds.

If original records are later obtained, outcome verification can be carried out retroactively against the original window, noting at what time the basis was acquired. Past outcomes may have occurred long ago; Gu Ning has learned only now. Retroactive adjudication does not automatically endow the original predictive version with the later material.

Loss of Contact Is Not a Complete Definition of Outcome

Loss of contact can mean that during a certain period one could not reach the other party, or it may mean that a team member does not know where to continue looking for material. Without specifying the object, the time window, and the channel, it cannot bear the negation of the original event or the final failure of cooperation on its own.

Consider a channel comparison: the person responsible for communication is temporarily unable to reach the other party, but the agreed channel may have received a message earlier; if the material cannot be returned, the state should be "awaiting verification," not rewritten as "message received" or "message not received" simply because contact is currently lost.

If the window has closed, the necessary communication material can be verified, and full confirmation was not received, the outcome can be marked as not held under the target, with the contact state recorded separately. Loss of contact may provide a process clue, but it does not add an undeclared negation condition to the definition of the event.

The main line of this book has not introduced new facts of loss of contact. The comparison is meant to let the receiver know what to ask when encountering this label in the future; one must not automatically interpret the four missing original records as the client's loss of contact at the time, nor deduce personal motives from a missing file.

Withdrawal Must State From What Process One Has Withdrawn

A request that ceases to continue repairs may be called a withdrawal from the cooperative process; if the predictive object is receipt of full confirmation within the original window, it is still possible that confirmation was first obtained and withdrawal then occurred. The fact of withdrawal and the target outcome can be preserved simultaneously, without needing to choose one to overwrite the other.

If the continuation is explicitly terminated before full confirmation, one must return to the target and window to check what it supports. The description of termination may be important, but whether there is still a possibility of later reopening or changed reply cannot be filled in by Gu Ning as "forever impossible" on the basis of impressions.

Another kind of withdrawal occurs within the observation system: a team member stops tracking the request, material becomes difficult to access after migration, or the registration is moved to archives. An object exiting the current search entrance does not mean that real-world cooperation was simultaneously terminated; the two processes must be separated.

The state table can therefore preserve three layers: business continuation, observational continuation, and target adjudication. They are related but are still not the same column. Otherwise, withdrawal easily becomes an ostensibly well-explained general label that in fact omits the original result.

Counting Only Completed Objects Leaves What Out

Consider another independent comparison of ten items, not added to the Chengwan registration. Suppose they all use the same relative window but entered at different times. On the day of organizing, six have windows that have closed and whose material is adjudicable: four held and two not held; the other four have windows that have not yet closed and have not received full confirmation.

Among the six completed items, the positive proportion is two thirds. Among the current ten items, the known positives are two fifths—but this figure is not the eventual window confirmation rate for all ten, because the four items may still yield later outcomes. The two numbers have different denominators and different uses.

If the four uncompleted items are deleted and one then says ten items generally show two thirds confirmation, the report has overshot the completed range. If they are marked negative and one then says only two fifths of the ten are confirmed, the future window blank has been converted into a known negation.

Completed items can serve as a clearly defined subset, but one must still check whether completion and entry time, task conditions, and the like are related before discussing migration to the whole. Waiting for later outcomes also cannot override the earlier time of each item; one does not pretend the observation windows are equally long merely because the organizing date is the same.

Window Expired but Material Missing—Another Counting Gap

Consider another independent comparison of twelve items, all explicitly using the same target, with all windows closed. Nine have adjudicable material: six held and three not held; three outcomes are temporarily unknown. The unknown results from insufficient outcome material, not from the observation window having not yet arrived.

Among the nine adjudicated items, the proportion is two thirds. If the goal is the actual proportion over all twelve, a complete figure has not yet been obtained. Only within this explicitly hypothetical table can the unknown three yield a number of eventual positives ranging from zero to three, so the total number of positives ranges from six to nine and the proportion ranges from one half to three quarters.

This range is a logical bound set by the known count and the binary target; it is not a statistical confidence interval, nor does it assign equal probability to each unknown object. It does not say the two ends are equally likely, and certainly does not guarantee anything about Chengwan's actual R17 outcome.

This method cannot be applied directly to the original sixteen items. Chengwan's other four differ in stage or window, whether all sixteen can be adjudicated against the same current target has not been established, and the eight-item set of the original five confirmations differs from the nine-item comparison. Bounds require their own complete setting and are not automatically calculated from missing a few objects.

Imputing Half to Each Unknown Does Not Yield a New Result

In the twelve-item hypothetical table, if the three unknown items are each filled in at half, a calculation of 7.5 positive items is produced. This can represent an expected value within another explicitly defined probabilistic model, but it is not an observation of 7.5 actual positive results.

The actual number of binary outcomes held should be an integer; the unknown true results do not become half-held through imputation. If a probabilistic projection is required, one must state how the unknown objects are estimated, why the same probability is adopted, and whether this is related to the missing material.

Merely because "unknown" sounds as if either side is possible, one fills in one half, repeating the equiprobable misunderstanding of Chapter 5. That three items are all unknown does not guarantee they share the same conditions; the blank label only says that adjudication has not yet been obtained, and does not provide a common distribution.

Gu Ning can preserve the original unknown state while also recording an explicit hypothetical projection and its use. Expected quantities and actual results must be separated so the receiver can update when later material arrives, without disguising the number previously filled in as fact.

Not in the Display Can Still Preserve a Place in the Outcomes

Gu Ning does not need to produce an achievement card for every incomplete request before allowing it to remain in the predictive material. She can preserve, under the original request identifier, the target version, the window, the current observational state, and the basis of the outcome—even without a photo or a permission.

The display entrance continues to operate for its own purpose; outcome preservation has a separate entrance. This way the achievement cards can introduce the facts of completion while the prediction ledger does not depend on whether display is ultimately possible. Absence of a card does not automatically delete the object, and addition of a card does not automatically increase the count of original-window positives.

If material for a certain item is temporarily missing, its identifier and the fields requiring verification are still preserved. If material is later obtained, there is no need to create a new request from scratch; if it is never obtained, the uncovered range can still be stated. Preserving the position does not impute the result; position and result can exist separately.

What this practice adds is continuity for the unobserved state, without requiring all history to have been fully checked. Chapter 19 will organize cross-batch prediction ledgers; this chapter first ensures that unknown objects do not lose their return entrance simply for lacking a final story.

States Should Not Rest on a Single General Label

At 3 p.m. on Thursday, Gu Ning separated the states in the original records table. The registration describes which objects entered the search range, the event adjudication states whether the target has a basis, the observation window states whether the opportunity has closed, material states which part of the original record is needed, and later business outcomes are recorded separately.

She did not fill in more false facts for unknown objects; instead, she reduced the meaning borne by a single field. R13–R16 remain as original records not obtained, current target outcomes not yet verified; R09–R12 remain as having material but differing in stage or window, not folded into the eight-item current frequency.

R01–R08 retain five held and three not held, with the explanation that outcomes are within the preserved original window. R17 remains as currently unconfirmed, original window not yet closed, reception time awaiting verification, not merged into the same category of "unsuccessful" as the four historical missing files.

Once the fields are separated, new material can take effect on specific states. Finding the original document changes material acquisition; window arrival changes the observational boundary; receipt of full confirmation changes the known state of the event; later completion changes the business outcome; a single change does not have to rewrite every column at once.

Outcome Verification and Narrative Closure Are Separate

When a story is written to a close, it may be because cooperation is complete, or because there is no more material to narrate. The narrator's stopping point does not automatically become the prediction window's deadline, and certainly does not automatically state that the target has received affirmation or negation.

If a request's story ends with no further contact between the parties, one must identify which time window it describes and whether the source is sufficient for adjudication. If the story ends with achievement photographs, one must also return to the original first-version confirmation time and not fill in "yes" for the past window on the strength of completion.

Narration can preserve process and meaning, but it does not by itself take over event adjudication. Chapter 17 will examine how stories cover the original prediction; here, state preservation is first given its own boundary so that the end of narration does not delete fields that remain unobserved.

Gu Ning does not demand that every story be traced to eternity. A finite stop can state the reason for stopping and the location of what remains unknown, without claiming that reality is known to have no further sequence. Ending the current search and ending the object's possible process are two different levels of judgment.

How Late-Arriving Results Are Filled Back In

If a missing request's original communication is found in the future, first verify whether it corresponds to the current target, version, and window, then record the outcome and the time of acquisition. If only a later completion photograph is found, it may fill in the fact of business completion but still cannot adjudicate original-window confirmation.

Imputing a result should return to the original object; one must not create a new request, nor write Gu Ning's later-acquired result into the information that was originally available. The version previously marked unknown is still preserved; the later version may reduce the material gap; each has its own basis in place.

If the object whose result is imputed changes the original scope of analysis, the new denominator and its reason must be stated. One cannot add the new positives to the numerator while omitting the failures or still-unknown objects; nor may one refuse to re-admit an object simply because its imputed result was unfavorable.

State preservation lets revision occur concretely. It does not guarantee that the probability will fall or rise once material is recovered, but it allows the result to enter under the same rule, preventing only the final survivor story from having the chance to change a judgment.

Internal Pauses Cannot Enter the Outcome Numerator Directly

A repair station may place a request on temporary hold; this is an internal continuity state. If the original predictive object is still receipt of full confirmation within the window, a temporary hold does not mean the client has refused, nor does it mean the agreed channel will not receive a definite reply afterward.

If a later internal pause changes the mode of communication, this may affect how the target occurs; what must be preserved is the action and the actual material, not a negative mark based on the pause label alone. Earlier, this book distinguished probability from action; this chapter applies the same division to the outcome column: the management state describes how the station handles things, and the target state describes how the event is observed.

Consider a comparison: after an internal pause, full first-version confirmation is still received within the window—both layers of state can hold simultaneously. Alternatively, preparation continues internally but no confirmation is received within the window—both layers can again hold simultaneously. They are not necessarily consistent, and the internal process state must not be used directly as an event adjudication.

These comparisons do not indicate that R17 has been paused or is being actively pursued. The state fields merely preserve the positions of different processes and do not generate action facts on behalf of the repair station. Only when the field meanings are clear is there an opportunity later to check how internal actions connect to predictive outcomes.

"Unknown" Also Cannot Be Used to Revoke a Known Negation

Preserving unknown states carefully does not mean that when an unfavorable outcome is seen, one says there is still a sequel and therefore the item cannot be counted as negative. The three original failures among the eight items are supported by window material; a later renewed cooperation does not erase the original target adjudication. An open process does not erase a clear time boundary.

If someone retains the five positives but changes the three failures to "still possible in the future," the table looks more optimistic, but in fact it mixes the receipt event with eventual cooperation. Positives are governed by the original window, while negatives are governed by an unlimited future; the two sides are no longer constrained by the same rule.

When a judgment needs changing, one should point to which original record does not correspond or what correction has been obtained, rather than merely because the ending can still change. Unknown is a state of insufficient basis, not a rhetorical device for protecting the model from unfavorable results.

This keeps observation open to both sides. Positives with sufficient support and negatives are both recorded; items with genuinely insufficient material also preserve "unknown." The rigor of the system is not bypassed through the abuse of caution, so that only smooth stories can remain in the outcome table.

How Gaps in Relations Continue to Be Checked

If material gaps are later found to be concentrated in a certain organizational process, that process can first be checked to see which objects it covers and why documents were hard to retrieve. The gap may be related to the target outcome, or it may mainly reflect how material was stored; the same object's material is required, and concentrated gaps cannot be used to prove hidden failures directly.

R13–R16 do not yet have this kind of provenance account; this chapter leaves only verification questions. If portions of the registration fields have been temporarily obtained, new fields may be preserved, but it cannot be said that the outcome has also been completed. One gap's shrinking does not guarantee that other gaps will vanish in turn.

If nothing is ever obtained, one can continue using the explicit description of the comparable eight items while noting the coverage limitation. Using a limited set does not mean claiming all material is complete, nor should the one-half historical frequency established over the verified eight items be revoked because the other four are missing; that number must defend its own denominator and not substitute for all outcomes.

Gaps in relation affect how migration proceeds; further models and material are needed. Currently there is no basis for assigning a fixed correction value. Preserving this interface is more helpful for later genuine verification than either imputing all unknowns to failure or deleting them entirely.

A Pending Outcome Is Not an Inability to Make Probabilistic Judgments

R17's outcome is undetermined—which is exactly the situation a probabilistic judgment must face. Gu Ning may form a candidate judgment from explicit material and assumptions, without waiting for complete confirmation before prediction is permitted; awaiting outcome adjudication and permitting limited estimation are two different tasks.

But she cannot write a candidate probability as if it were the result, nor can she fabricate an original prediction she never actually adopted just because she will calibrate it in the future. The probabilistic version preserves the information at the time; the result version preserves the later basis; the two can correspond without impersonating each other.

Part Four will handle the comparison between predictive groups and outcomes. This chapter first ensures that not-yet-adjudicated objects still have a recording position: only when the later window and material are both complete can they be checked against the same event; if material is still insufficient, one can state why the item cannot yet enter a given outcome tally.

Each state change should also note who verified it and on what basis, so the receiver knows the change came from new material, window arrival, or an explanatory correction. Unchanged fields continue to preserve their original basis, and a final saved date cannot be used in place of all observation times. Records may be brief, but they must be sufficient to trace back to the adjudication—not merely leave behind a new label that nobody can explain.

How Finite Observation Is Delivered

At 3:20 p.m. on Thursday, Gu Ning preserved the state account. The original eight items' window outcomes are maintained; the four items with different stage or window are kept separately for other uses; the four missing original records remain unadjudicated; R17 is currently unconfirmed and its window not yet closed. The eight display cards do not substitute for the outcome column of any of these sets.

She preserves each state's verification interface, refusing to lump loss of contact, withdrawal, no card, and incomplete into "failure," nor to delete objects merely because an outcome is temporarily unobtainable. Two independent hypothetical counts explain respectively the unobserved-but-window-open and the completed-but-material-missing, without being folded into the Chengwan registration.

RC's processual completeness is realized as a representation that admits continuation when new material enters; the unknown position does not vanish for want of a story. Finite observation does not exhaust an object's process, nor does it imply that all states are equally unknowable; the adjudicated and the not yet adjudicated are each preserved according to actual material.

The judgment reached in this chapter is that unobserved outcomes first require identifying the time window, the material, and the object's state. Known negatives can be counted; unknowns must not be arbitrarily filled in; a later ending does not adjudicate the original window. Judgments may be limited in scope, but delivery must state what has not yet been obtained.

The next chapter turns to materials that covary. Even where target outcomes are all recorded, the co-occurrence of two conditions does not by itself say which caused which. State preservation lets objects enter comparison; explanation still needs its own conditions, and a complete table must not be allowed to make causation hold prematurely.