What's new in SimuXL

A comparison prints a change from the Base case in percentage points, calls its first chart an exceedance curve when percentiles are read Above, and never draws another alternative in the Base case’s colour; a figure SimuXL computes is never printed as 0 when it is not 0; and a run ends with your workbook at rest. 4.16.3 follows 4.16.1 and carries the change prepared as 4.16.2, which was never released on its own. Every change is a fix: no command or function is added or removed, the trials a run draws are unchanged, the license agreement is unchanged, and SimuXL is signed with 4.14.0’s certificate.

Fixed

• A change from the Base case is printed in points. The comparison sheet’s two-targets block prints each alternative’s change from the Base case — both met, and each target met — as a difference of two shares. It printed that difference as a percent, so a fall from 50.7% to 41.2% read -9.5%, which reads as a relative change, while the note under the table said percentage points. The cell now holds the points and prints them with their unit — -9.5 pts — on the sheet and in both comparison reports.
• Under Above, the comparison’s first chart is called an exceedance curve. When a workbook reads percentiles as Above, the first chart of each output plots the share of trials at or above each value, so each curve falls from left to right. Its foot, the comparison report’s caption and the sheet’s note when the chart cannot be drawn called it a cumulative distribution; they now call it an exceedance curve. Under Below nothing changes.
• No alternative is drawn in the Base case’s colour. A comparison draws its alternatives in five colours, and a sixth alternative took the first colour — usually the Base case’s — so two curves on the chart could not be told apart. Once the five are used, the colours now cycle over the four that are not the Base case’s, wherever the Base case sits.
• A figure SimuXL computed is never printed as 0 when it is not 0. A mean, a standard deviation, a percentile or a band that the output’s own number format would show as 0 — a count’s mean of 0.0752 under a whole-number format — printed as 0, so SXL Convergence said The mean of 0 is settled to ±0.010 … which is 13.61% of the mean. Such a figure is now printed to two significant figures in the same format — 0.075 — on SXL Convergence and in its sentences, in the statistics panel and the Output Viewer, on the SXL Histogram sheet and its charts’ markers, in the run’s output table, at the Tail Drivers threshold, on the comparison sheet and in the reports. A true zero is still printed 0, and the Base — your cell’s own value at rest — is printed as your cell shows it.
• A run ends with your workbook at rest (prepared as 4.16.2). In Manual calculation a run left the cells its trials moved showing the last trial — your model on screen, a workbook linked to it and the SXL.Live cells — and in Automatic except for data tables it left the data tables over your outputs on it. The Histogram command, the Output Viewer and the run’s reports, which read each output’s base value from the workbook when you use them, then drew that trial’s value as a base. SimuXL now recalculates every open workbook once when a run or a comparison ends — finished, stopped, or stopped by an error — in any calculation mode but Automatic, and once after the Impact analysis’s swing, however it ended. If that recalculation fails, SimuXL says so, and Ctrl+Alt+F9 recalculates the workbook.

Help and the User Guide

• The User Guide says the comparison’s change cells hold percentage points and what a formula reading one gets; names the exceedance curve under Above; says a figure SimuXL computes is never printed as 0 when it is not 0; and says a run ends with your workbook at rest.

A run’s record names the control variables it used, a Latin Hypercube run’s mean standard error comes from its designs, the Tornado’s shares are never negative and the Tornado says where they overlap, and the run report lists the distributions each fit compared. 4.16.1 follows 4.15.0 and carries the changes prepared as 4.15.1 and 4.16.0, which were never released on their own; 4.16.1’s own changes are fixes. The trials a run draws are unchanged, and three figures computed from them change, as described below: a Latin Hypercube run’s mean standard error, the Tornado’s shares, and Convergence’s comparison of the two errors. No command or function is added or removed, the license agreement is unchanged, and SimuXL is signed with 4.14.0’s certificate.

New

• The run record names every control variable and the value it held. A run’s record — the SXL Trial Run sheet and the run’s HTML and Word reports — has a Control variables row: each control variable’s name, its cell and the value its cell held during the run, as the cell showed it, such as Real discount rate (Inputs!$G$15) = 3.5%. A workbook with none says so; a cell that calls SXL.Control but is not taken as a control variable is listed too, with its value and a note that it is not; and where SimuXL could not read a sheet, the row names the sheet. The Random seed row, which says the same seed and workbook reproduce the run exactly, now adds that this holds with each cell at the value the record lists — and any SimuXL could not read at the values they held.
• The run report lists the distributions each fit compared. In the HTML and Word reports, the table of inputs fitted to data or to percentiles has a column Families compared (A-D): for a fit to data, every distribution it tried, with its Anderson-Darling statistic — lower is closer — and the one written into the input marked, such as Normal 0.279 (chosen); Lognormal 0.464; Uniform 1.992; Triangular 0.337; Beta-PERT 0.375. A fit to percentiles, which compares no families, shows a dash, and so does a fit whose input cell no longer holds the formula the fit wrote.

Changed

• A Latin Hypercube run’s mean standard error comes from its designs. The statistics panel’s Mean std. error — on the SXL Histogram sheet, in the Output Viewer and in the reports — was the standard deviation over the square root of the trials on every run. That assumes independent trials, which a Latin Hypercube run is built not to have, and on such a run it could overstate the error many times over: a cost whose designs put its mean’s standard error at $2,716 printed $44,793. On a Latin Hypercube run it is now the spread of the designs’ means over the square root of their count — the figure SXL Convergence measures — and its label says how many designs: Mean std. error (20 designs). Where Convergence cannot measure the designs, the panel prints a dash and its label says why. A plain Monte Carlo run is unchanged.
• The Tornado’s shares are never negative, and it says where they overlap. Contribution to Rank Variance was the squared rank correlation with its sign, so an input that moves an output the other way printed a negative share. It is now each input’s squared rank correlation as a share of all of them; the ρ beside it gives the direction, and the Contribution picture’s bars all point one way. Where inputs share an effect — two ranked inputs correlated with each other, or an input ranked as itself that is also inside a made input — a Shares overlap row under that output says so and names the correlated pairs among the rows its table shows. A pair both of whose shares are under 0.05% is counted there rather than named: 3 other correlated pairs among these rows are not named: both shares of each are under 0.05%. A pair one of whose shares prints is still named, because an input can drive the output while its own share is near zero, its partner’s rank correlation carrying its effect. The row is on the SXL Tornado sheet, in the Output Viewer and in the reports, and an output where nothing is left to name has none.

Fixed

• The SXL Histogram a run writes draws each output’s base value. It was drawn without the settings made under the Output Viewer’s Configure this chart — each output’s base value, markers, split and colours — so it printed the base value as “-” while the Histogram command, the Output Viewer and the reports drew the same run with them. The run now reads them once, before its first trial, while the workbook is at rest, and its Base row says so; if they cannot be read, the sheet and the run’s completion window say so.
• The Output Viewer’s statistics panel shows its labels and values whole. Labels such as Mean std. error (20 designs), Std deviation and Mode (tallest bin), and some values, were cut short. Each column is now as wide as what it holds, and the panel takes width from the chart, up to half the window, then scrolls sideways rather than cutting a word. The viewer can no longer be made narrower than its top row, where the Statistics box sat over the Output label.
• Saving Preferences leaves you where you were. The first save into a workbook — Save in Preferences or Report Preferences, in most workbooks — added SimuXL’s hidden sheet and moved you to the workbook’s last sheet. You now stay on the sheet and the cells you had selected; the hidden sheet Convert keeps its undo record on is added the same way.
• Convergence’s two estimates of the mean’s error divide as printed. The sentence comparing the textbook error with the designs’ own error printed two 95% bands, with different multipliers, beside a multiple that was a ratio of variances, so the bands divided to a different number. It now prints the two standard errors, each to at least three significant figures where the output’s format can show them, how many times smaller the second is, and how many independent trials the run is as precise as — $44,793 … and $2,716 … about 16.5 times smaller - as precise as about 2,700,000 independent trials — and the sampling error the multiple and that count carry.
• Changing an input’s Type writes that Type’s function. Define Input, opened on a cell holding one distribution with a different Type chosen, wrote the old function’s name around the new parameters — a Triangular changed to a Normal came back as =SXL.Triangular(1.07, 0.0996). It now writes =SXL.Normal(1.07, 0.0996).
• A band too small for its output’s format is printed, not shown as zero. Convergence printed a band under half its format’s last digit as zero — settled to ±0.0 … which is 0% of the mean. Such a band, the figures behind it, its share of its figure and the statistics panel’s mean standard error are now carried to two significant figures in the same format, on the sheet, in the reports and in the panel; a date, fraction or scientific format is left as it was.

Help and the User Guide

• The User Guide’s Tornado section says the shares are never negative and which pairs the Shares overlap row names and counts; its Convergence section says how the two standard errors are compared and that a Latin Hypercube run’s panel gives the between-design figure; and its seed paragraph says a run reproduces with every control variable at the value the record lists.

Two things are added to Joint Outcome — the joint-confidence frontier, and a reading of what it counted — and its chart draws at most 3,000 trials and says how many. Nothing a run computes changes, and no command or function is removed. The license agreement is unchanged, and SimuXL is signed with 4.14.0’s certificate.

New

• Joint Outcome draws the joint-confidence frontier. Under the second output, the dialog has a new tick, Draw the joint-confidence frontier at, and a level in percent. It starts ticked, at the workbook’s Shade either side of under Preferences › Charts when that is above 0 and below 100, otherwise at 70; in one Excel session the dialog keeps the tick as you last left it and a level you typed. For each threshold of the first output, SimuXL finds the lowest threshold of the second at which both are still met in at least that share of the trials — for a date, the earliest day; for an output met at or above its threshold, the highest, or the latest day. Pressing Joint Outcome with a row’s two thresholds counts both met in at least that share. The chart draws the frontier as a purple line; the SXL Joint Outcome sheet gets a JOINT-CONFIDENCE FRONTIER table, with how many trials meet each row’s first threshold, and sentences saying where the frontier starts, where it stops moving and how steady a row is; the message box says where it runs; and the HTML and Word reports carry the table and the line.
• Joint Outcome writes a reading of what it counted. Above the chart the sheet has a READING block — findings written by rules from every counted trial, then a bottom line — and the reports carry the same sentences. It says how often each output meets its threshold and how often both do; which is the harder to meet, or that neither clearly is; how both met compares with multiplying the two shares; and the two outputs’ rank correlation, whose words about meeting the two together are given only where the counts at your thresholds lean the same way. With the frontier drawn, it says whether your thresholds already give both at its level and, if not, holding each of yours in turn, the other threshold with which both reach it and how far it lies from yours — or that none does; and it reads the frontier at the first output’s own points above its level (75%, 80%, 90% and 95% for a 70% frontier), leaving out a step that buys nothing. For two numbers met the same way in the same number format, one line says each is counted against its own threshold. It names up to three inputs the harder output moves most with — each output’s when neither is the harder — from the same ranking as the Tornado. The words say what the trials show, never why, and give no advice about your project.

Changed

• Joint Outcome’s chart draws at most 3,000 trials, and says how many. Beyond 3,000 it draws 3,000, spread evenly through the run, where it drew up to 20,000, so the cloud stays readable; its legend reads, for example, Trials - 3,000 of 10,000 shown (30.0%), and the sheet and the report say the same. Every count, the four outcomes’ labels, the frontier and the reading still use every trial.
• What Input refuses is said as it is. The tip on Input, the User Guide and Define Input’s refusal said an input cell must hold a plain number. Define Input refuses a cell holding one of your model’s formulas, because a distribution written over it would break the model; an empty cell, a number or text opens the dialog, and so does a cell holding one SimuXL distribution on its own, to edit it. The tip and the refusal now say that a model’s formula is refused, and the User Guide says which cells open the dialog; a cell with two or more distributions, or one inside a larger formula, is still refused — edit those by typing.

Fixed

• A Joint Outcome threshold of an output shown in thousands or millions, such as #,##0.000,,, is printed in full where the format would round it — on the sheet, in the log line, in the reports and in a comparison’s targets: a threshold of 1,234,567 had printed as 1.235.

Help and the User Guide

• The User Guide’s Joint Outcome section describes the frontier, the reading and the thinned chart, and its Input section says which cells open Define Input.

The license agreement moves into the add-in, SimuXL is signed in its publisher’s name, Simuraytics Software LLC, and the Alternatives button opens a dialog that sets up your alternatives. 4.14.0 follows 4.9.2 and carries the changes prepared as 4.10.0 to 4.13.0, which were never released on their own. Apart from two changes below — where a mitigation cost goes, and which analyses a run builds until you save your Preferences — nothing a run computes changes, and no command or function is removed.

New

• The license agreement is in the add-in. The first SimuXL command you use on a machine shows the SimuXL Software License Agreement, once, with an I agree box, and Help › License shows it at any time. Activate works only once the box is ticked; the trial runs either way. SimuXL keeps the record of what it showed, and of your tick, in a small file beside its log, SimuXL.agreement; Help › About has an Agreement line saying whether the box was ticked on this machine; and Help › License has a button for the third-party notices.
• Alternatives opens a set-up dialog. One dialog for the whole Alternative Table: add and remove alternatives, give each a name and a description, and change the control values it changes — a new alternative’s values start at the base case. OK checks every name and value first, then writes only the cells it changes on the SXL Alternatives sheet, never over a formula; Cancel writes nothing. Drag an edge or the grip in the lower right corner to make the dialog larger. The User Guide walks through it.
• A formula from the … button in Define Input. The … beside each parameter takes a formula such as =B4*0.9 as well as a click on a cell, and writes every cell it names as a relative reference, so an input copied down a column follows its own row. A $ reference you want kept, type into the parameter’s box itself.
• What a percentile input was entered as. For an input entered by the percentiles you know, such as =SXL.Triangular(80, 100, 130,, "P10,P50,P90"), Find Inputs has a new Entered as column — P10 80; P50 100; P90 130 — as does SXL.HarvestInputsArray for VBA, and the Input Summary and the HTML and Word reports say the same beside the parameters SimuXL solved from it.
• A fit follows its input. A fit you use now keeps a hidden name for its input’s cell (SXL_Fit_0001 and on), so Find Inputs, the Input Summary and the reports still name the fit after you insert rows above the input, cut and paste it, or rename its sheet; and the fit’s record on the SXL Fit sheet is rewritten to say where its input now is.

Changed

• Mitigation costs go in your model. The Alternative Table’s Mitigation cost row, Run Comparison’s Mitigation cost list, and the MITIGATION COST block on the comparison sheet and in a comparison’s report are gone. Give a cost a control variable in your model — =SXL.Control("Mitigation cost", 0) in an empty cell, added into the formula of the output it costs — and set each alternative’s cost in Alternatives. Your own formulas then escalate, mark up or vary it wherever they do that to your other costs. The User Guide’s A mitigation cost in your model has a worked example.
• All five analyses until you save your Preferences. In a workbook whose Preferences have never been saved — a new one, or one you already have — Tornado, Impact, Tail Drivers and the Input Summary start ticked under Preferences › Analyses, as the Histogram does, so the next run builds all five. A workbook whose Preferences were saved keeps the ticks it was saved with. To keep one off, untick it and press Save.
• Simuraytics Software LLC. SimuXL’s publisher and licensor is Simuraytics Software LLC, the new name of MS Software Services LLC. The add-in and the installer are signed in its name; the installer, Windows’ Installed apps and Help › About name it; and the installer now offers the trust step ticked on every install.
• License Agreement Version 1.2, effective 2026-10-15. It names the licensor by its new name; says that activating a code sends SimuXL’s version number with the code and the machine id; gives simuraytics@gmail.com and 404-997-3543 for support; and rewrites Section 6, Refunds: a new license or subscription is refunded in full on a request by email, with the receipt number, within 30 days of its purchase, and extensions and renewals are not refundable, except for a duplicate or erroneous charge.
• What the update check sends. The update dialog, its button’s tip, the weekly check’s note and the User Guide now say what the check’s request carries: SimuXL’s version number, and no machine id, licensee, key or workbook name — though, like any web request, it reaches the web host with your IP address. The check itself is unchanged.
• Impact’s estimate of its own cost. Impact’s tip, its note in View Results and the User Guide now give the estimate SimuXL quotes before a long Impact starts: roughly 0.2 seconds at 100 inputs and 20 seconds at 1,000.
• Event macros. SimuXL writes its hidden sheet and names, and Export Trials writes its workbook, with Excel’s events off, so an event macro such as Workbook_SheetChange no longer runs for each block of cells SimuXL writes there.

Fixed

• A parameter box in Define Input holding only = is refused, as an empty box is, instead of being written as 0.
• Run Comparison’s note for two or more ignored rows reads “this workbook does not have” and “on them”, where it said “do not have” and “on it”.
• In Fit…, the status line under the percentile boxes carries its whole text in a tip, as the lines on the data side do.

Help and the User Guide

• Each parameter box’s tip says a box takes a formula and that it is worked out every time the cell calculates, and the formula help for each distribution’s Parameters argument says what its numbers are, inside the 255 characters Excel shows.
• The User Guide’s confidence example uses 70% and P70, the level the highway industry uses.
• The User Guide says SimuXL works offline and sends nothing about your work, and that activating a code sends the code, the machine id and SimuXL’s version number.
• The User Guide no longer carries notes about earlier releases.